Search

Introduction Artificial Intelligence (AI) chatbots are able to explain complex concepts using plain language. The aim of this study was to assess the accuracy of three AI chatbots answering common questions related to contact lens (CL) wear. Methods: Three open access AI chatbots were compared: Perplexity, Open Assistant and ChatGPT 3.5. Ten general CL questions were asked to all AI chatbots on the same day in two different countries, with the questions asked in Spanish from Spain and in English from the U.K. Two independent optometrists with experience working in each country assessed the accuracy of the answers provided. Also, the AI chatbots’ responses were assessed if their outputs showed any bias towards (or against) any eye care professional (ECP). Results: The answers obtained by the same AI chatbots were different in Spain and the U.K. Also, statistically significant differences were found between the AI chatbots for accuracy. In the U.K., ChatGPT 3.5 was the most and Open Assistant least accurate (p < 0.01). In Spain, Perplexity and ChatGPT were statistically more accurate than Open Assistant (p < 0.01). All the AI chatbots presented bias, except ChatGPT 3.5 in Spain. Conclusions: AI chatbots do not always consider local CL legislation, and their accuracy seems to be dependent on the language used to interact with them. Hence, at this time, although some AI chatbots might be a good source of information for general CL related questions, they cannot replace an ECP

[Abstract]: Optical coherence tomography angiography (OCTA) is a non-invasive imaging modality used to evaluate the retinal microvasculature. Recent advances in OCTA allows to visualize the blood flow within the choriocapillaris region, where a granular image is obtained showing a pattern of small dark regions, called flow voids (FVs). Given its relevance, numerous clinical studies have linked the changes in FVs distribution to multiple diseases. The granular structure of these images makes accurate labeling and segmentation difficult, which can be overcome by using a multi-target perspective. However, manually designing a neural architecture that can accurately predict all targets in a balanced way is a major challenge. In this work, we propose a novel methodology based on evolutionary multi-target optimized networks that, through a set of evolutionary operators, traverses a search space of architectures in a deep but efficient way. This methodology allows us to discover efficient and accurate multi-target architectures tailored to our problem, but which are also adaptable to other tasks due to their robustness. To validate and analyze our methodology and the discovered network model, we performed extensive experimentation with cases from a real clinical study, achieving better results than the state of the art and manually designed architectures.

Purpose: To evaluate the glistening in 4 different models of intraocular lenses (IOLs) using optical coherence tomography (OCT) and deep learning (DL). Setting: Centro Internacional de Oftalmología Avanzada (Madrid, Spain). Design: Cross-sectional study. Methods: 325 eyes were assessed for the presence and severity of glistening in 4 IOL models: ReSTOR+3 SN6AD1 (n = 41), SN60WF (n = 110), PanOptix TFNT (n = 128) and Vivity DFT015 (n = 46). The presence of glistening was analyzed using OCT, identifying the presence of hyperreflective foci (HRF) in the central area of the IOL. A manual and an original DL-based quantification algorithm designed for this purpose was applied. Results: Glistening was detected in 22 (53.7%) ReSTOR SN6AD1, 44 (40%) SN60WF, 49 (38.3%) PanOptix TFNT, and 4 (8.7%) Vivity DFT015 IOLs, when any grade was considered. In the comparison of the different types of IOLs, global glistening measured as total HRF was 17.3 ± 25.9 for the ReSTOR+3; 9.3 ± 15.7 for the SN60WF; 6.9 ± 10.5 for the PanOptix; and 1.2 ± 2.6 for the Vivity (P < .05). There was excellent agreement between manual and DL-based quantification (≥0.829). Conclusions: It is possible to quantify, classify and compare the glistening severity in different IOL models using OCT images in a simple and objective manner with a DL algorithm. In the comparative study, the Vivity presented the lowest severity of glistening.

[Abstract]: Multi-task learning is a promising paradigm to leverage task interrelations during the training of deep neural networks. A key challenge in the training of multi-task networks is to adequately balance the complementary supervisory signals of multiple tasks. In that regard, although several task-balancing approaches have been proposed, they are usually limited by the use of per-task weighting schemes and do not completely address the uneven contribution of the different tasks to the network training. In contrast to classical approaches, we propose a novel Multi-Adaptive Optimization (MAO) strategy that dynamically adjusts the contribution of each task to the training of each individual parameter in the network. This automatically produces a balanced learning across tasks and across parameters, throughout the whole training and for any number of tasks. To validate our proposal, we perform comparative experiments on real-world datasets for computer vision, considering different experimental settings. These experiments allow us to analyze the performance obtained in several multi-task scenarios along with the learning balance across tasks, network layers and training steps. The results demonstrate that MAO outperforms previous task-balancing alternatives. Additionally, the performed analyses provide insights that allow us to comprehend the advantages of this novel approach for multi-task learning.

[Absctract]: Background: The COVID-19 can cause long-term symptoms in the patients after they overcome the disease. Given that this disease mainly damages the respiratory system, these symptoms are often related with breathing problems that can be caused by an affected diaphragm. The diaphragmatic function can be assessed with imaging modalities like computerized tomography or chest X-ray. However, this process must be performed by expert clinicians with manual visual inspection. Moreover, during the pandemic, the clinicians were asked to prioritize the use of portable devices, preventing the risk of cross-contamination. Nevertheless, the captures of these devices are of a lower quality. Objectives: The automatic quantification of the diaphragmatic function can determine the damage of COVID-19 on each patient and assess their evolution during the recovery period, a task that could also be complemented with the lung segmentation. Methods: We propose a novel multi-task fully automatic methodology to simultaneously localize the position of the hemidiaphragms and to segment the lung boundaries with a convolutional architecture using portable chest X-ray images of COVID-19 patients. For that aim, the hemidiaphragms’ landmarks are located adapting the paradigm of heatmap regression. Results: The methodology is exhaustively validated with four analyses, achieving an 82.31% +- 2.78% of accuracy when localizing the hemidiaphragms’ landmarks and a Dice score of 0.9688 +- 0.0012 in lung segmentation. Conclusions: The results demonstrate that the model is able to perform both tasks simultaneously, being a helpful tool for clinicians despite the lower quality of the portable chest X-ray images.

[Absctract]: Respiratory diseases have a significant global impact, and assessing these conditions is crucial for improving patient outcomes. Chest X-ray is widely used for diagnosis, but expert evaluation can be challenging. Automatic computer-aided diagnosis methods can provide support for clinicians in these tasks. Deep learning has emerged as a set of algorithms with exceptional potential in such tasks. However, these algorithms require a vast amount of data, often scarce in medical imaging domains. In this work, a new data augmentation methodology based on adapted generative latent diffusion models is proposed to improve the performance of an automatic pathological screening in two high-impact scenarios: tuberculosis and lung nodules. The methodology is evaluated using three publicly available datasets, representative of real-world settings. An ablation study obtained the highest-performing image generation model configuration regarding the number of training steps. The results demonstrate that the novel set of generated images can improve the performance of the screening of these two highly relevant pathologies, obtaining an accuracy of 97.09%, 92.14% in each dataset of tuberculosis screening, respectively, and 82.19% in lung nodules. The proposal notably improves on previous image generation methods for data augmentation, highlighting the importance of the contribution in these critical public health challenges.

[Abstract]: Recent clinical studies have emphasized the importance of understanding the morphology and mechanics of the ciliary muscle. The ciliary muscle plays a vital role in various functions related to the anterior segment of the eye, including the regulation of intraocular pressure and the maintenance of the shape of the crystalline lens. To advance research in this area, we propose a fully automated methodology for the segmentation and biomarker measurement of the ciliary muscle in two different scan depths (6 mm and 16 mm), which are commonly used by clinicians to analyze biomarkers. Our methodology aims to provide repeatable, and immediate results through an exhaustive analysis of different network architectures, encoders, and transfer learning strategies. We also extracted a comprehensive set of relevant biomarkers, including parameters that provide essential information about its behavior during the accommodation process, overall dimensions, and biomechanical properties. These biomarkers can help clinicians and researchers in the diagnoses and monitor of different ocular diseases such as glaucoma, myopia, and presbyopia and develop new therapeutic strategies, potentially leading to more effective treatments and improved patient outcomes. Our methodology achieved accurate qualitative and quantitative results, with high accuracy values of 0.9665 ± 0.1280 and 0.9772 ± 0.0873 for the best combinations for 6 mm and 16 mm, respectively. Our proposed system provides a valuable and automatic tool for clinicians and researchers in the segmentation and analysis of the ciliary muscle in AS-OCT images.

[Absctract]: Central Serous Chorioretinopathy (CSC) is a retinal disorder caused by the accumulation of fluid, resulting in vision distortion. The diagnosis of this disease is typically performed through Optical Coherence Tomography (OCT) imaging, which displays any fluid buildup between the retinal layers. Currently, these fluid regions are manually detected by visual inspection a time-consuming and subjective process that can be prone to errors. A series of six deep learning-based automatic segmentation architectural configurations of different levels of complexity were trained and compared in order to determine the best model intended for the automatic segmentation of CSC-related lesions in OCT images. The best performing models were then evaluated in an external validation study. Furthermore, an intra- and inter-expert analysis was conducted in order to compare the manual segmentation performed by expert ophthalmologists with the automatic segmentation provided by the models. Test results of the best performing configuration achieved a mean Dice of in the internal dataset. In the external validation set, these models achieved a level of agreement with human experts of up to 0.960 in terms of Kappa coefficient, contrasting with a value of 0.951 for agreement between human experts. Overall, the models reached a better agreement with either of the human experts than these experts with each other, suggesting that automatic segmentation models for the detection of CSC-related lesions in OCT imaging can be useful tools for assessing this disease, reducing the workload of manual inspection and leading to a more robust and objective diagnosis method.

[Abstract]: Retinal vascular tortuosity is an excessive bending and twisting of the blood vessels in the retina that is associated with numerous health conditions. We propose a novel methodology for the automated assessment of the retinal vascular tortuosity from color fundus images. Our methodology takes into consideration several anatomical factors to weigh the importance of each individual blood vessel. First, we use deep neural networks to produce a robust extraction of the different anatomical structures. Then, the weighting coefficients that are required for the integration of the different anatomical factors are adjusted using evolutionary computation. Finally, the proposed methodology also provides visual representations that explain the contribution of each individual blood vessel to the predicted tortuosity, hence allowing us to understand the decisions of the model. We validate our proposal in a dataset of color fundus images providing a consensus ground truth as well as the annotations of five clinical experts. Our proposal outperforms previous automated methods and offers a performance that is comparable to that of the clinical experts. Therefore, our methodology demonstrates to be a viable alternative for the assessment of the retinal vascular tortuosity. This could facilitate the use of this biomarker in clinical practice and medical research.

Background: Glaucoma is the leading global cause of irreversible blindness. Glaucoma patients experience a progressive deterioration of the retinal nervous tissues that begins with a loss of peripheral vision. An early diagnosis is essential in order to prevent blindness. Ophthalmologists measure the deterioration caused by this disease by assessing the retinal layers in different regions of the eye, using different optical coherence tomography (OCT) scanning patterns to extract images, generating different views from multiple parts of the retina. These images are used to measure the thickness of the retinal layers in different regions. Methods: We present two approaches for the multi-region segmentation of the retinal layers in OCT images of glaucoma patients. These approaches can extract the relevant anatomical structures for glaucoma assessment from three different OCT scan patterns: circumpapillary circle scans, macular cube scans and optic disc (OD) radial scans. By employing transfer learning to take advantage of the visual patterns present in a related domain, these approaches use state-of-the-art segmentation modules to achieve a robust, fully automatic segmentation of the retinal layers. The first approach exploits inter-view similarities by using a single module to segment all of the scan patterns, considering them as a single domain. The second approach uses view-specific modules for the segmentation of each scan pattern, automatically detecting the suitable module to analyse each image. Results: The proposed approaches produced satisfactory results with the first approach achieving a dice coefficient of 0.85±0.06 and the second one 0.87±0.08 for all segmented layers. The first approach produced the best results for the radial scans. Concurrently, the view-specific second approach achieved the best results for the better represented circle and cube scan patterns. Conclusions: To the extent of our knowledge, this is the first proposal in the literature for, Ministerio de Ciencia, Innovación y Universidades (España), Consellería de Cultura, Educación, Formación Profesional e Universidades, Xunta de Galicia (España), Centro de Investigación de Galicia, Unidad Docente de Inmunología, Oftalmología y ORL, Fac. de Óptica y Optometría, TRUE, inpress

[Resumo]: Neste estudo, proponse unha actividade de innovación docente no contexto da materia Tecnoloxías da Información e da Comunicación (TIC) do Grao en Inglés da Universidade da Coruña (UDC), co obxectivo de promover e facilitar o desenvolvemento de competencias en linguaxe inclusiva e non sexista entre o estudantado. A actividade consiste na creación dun documento de texto en formato Open Document Text (ODT) formatado usando estilos en LibreOffice Writer, a partir dun conxunto de contidos sen formato e recursos relacionados coa linguaxe inclusiva. Esta proposta busca que o alumnado reflexione sobre o uso actual da linguaxe na sociedade, identifique contidos sexistas e comprenda as súas implicacións. Ademais, preténdese que os estudantes aprendan estratexias para un uso non sexista da linguaxe, tanto no plano gramatical como no discursivo. Para avaliar a percepción do estudantado respecto ao uso da linguaxe inclusiva, realizouse un estudo previo mediante unha enquisa anónima. Os resultados desta enquisa servirán como base para a implementación e mellora da actividade proposta. A través dun enfoque activo e participativo, o estudo busca fomentar a concienciación sobre a importancia da linguaxe inclusiva e desenvolver habilidades prácticas no manexo de ferramentas de procesado de texto, contribuíndo así ao avance cara unha sociedade máis igualitaria e inclusiva.

[Abstract]: This chapter presents a software tool for the automated segmentation of diffuse retinal thickening regions from optical coherence tomography (OCT) images. For this purpose, two retinal regions were defined and extracted: the inner retina and the outer retina. Then, a learning process was used to analyze a comprehensive and heterogeneous subset of relevant patterns in the OCT scans. Finally, two complementary post-processing stages were applied to improve the obtained performance and the overall efficiency of the presented tool. The presented tool achieved satisfactory performance, which demonstrates the suitability of the adopted solution.

[Absctract]: There are different retinal vascular imaging modalities widely used in clinical practice to diagnose different retinal pathologies. The joint analysis of these multimodal images is of increasing interest since each of them provides common and complementary visual information. However, if we want to facilitate the comparison of two images, obtained with different techniques and containing the same retinal region of interest, it will be necessary to make a previous registration of both images. Here, we present a weakly supervised deep learning methodology for robust deformable registration of multimodal retinal images, which is applied to implement a method for the registration of fluorescein angiography (FA) and optical coherence tomography angiography (OCTA) images. This methodology is strongly inspired by VoxelMorph, a general unsupervised deep learning framework of the state of the art for deformable registration of unimodal medical images. The method was evaluated in a public dataset with 172 pairs of FA and superficial plexus OCTA images. The degree of alignment of the common information (blood vessels) and preservation of the non-common information (image background) in the transformed image were measured using the Dice coefficient (DC) and zero-normalized cross-correlation (ZNCC), respectively. The average values of the mentioned metrics, including the standard deviations, were DC = 0.72 ± 0.10 and ZNCC = 0.82 ± 0.04. The time required to obtain each pair of registered images was 0.12 s. These results outperform rigid and deformable registration methods with which our method was compared.

[Abstract] Intraretinal fluid accumulation is both the common symptom and culprit of the main causes of blindness in developed countries: Age-related Macular Degeneration and Diabetic Macular Edema. For its diagnosis, experts of the domain employ Optical Coherence Tomography images (OCT), providing non-invasive cross-sectional representations of the retinal structures. However, like any medical imaging modality, OCT is influenced by multiple factors that impact its quality and subsequent interpretation. Coupled with the subjectiveness of the human experts, these factors can significantly affect the diagnostic process, treatment and quality of life for the affected individuals (particularly in these pathologies where early detection is crucial). To address these challenges, Computer-Aided Diagnosis (CAD) methodologies are developed, offering a layer of abstraction of the information present in the images. Still, in the particular scenario of these pathological fluid accumulations, the development of these methodologies is specially difficult due to their diffuse nature without defined boundaries. In this thesis, we proposed different CAD methodologies with the objective of helping expert clinicians to better detect and understand these pathologies. Furthermore, we expand the developed methodologies to other medical imaging modalities and conditions, such as macular neovascularizations in OCT Angiographies and COVID-19 diagnosis through the analysis of lung chest radiographs., [Resumen] La acumulación de líquido intrarretiniano es tanto síntoma común como culpable de las principales causas de ceguera en los países desarrollados: la degeneración macular asociada a la edad y el edema macular diabético. Para su diagnóstico, los expertos en el campo emplean imágenes de Tomografía de Coherencia Óptica (OCT), que proporcionan representaciones transversales no invasivas de las estructuras retinianas. Sin embargo, al igual que cualquier modalidad de imagen médica, OCT se ve influenciado por múltiples factores que afectan a su calidad y posterior interpretación. Junto con la subjetividad de los expertos humanos, estos factores pueden afectar significativamente el proceso diagnóstico, tratamiento y calidad de vida de las personas afectadas (particularmente en estas patologías donde una detección temprana es crucial). Para abordar estos desafíos, se desarrollan metodologías de diagnóstico asistido por ordenador (CAD), que ofrecen una capa de abstracción de la información presente en las imágenes. Sin embargo, en el escenario particular de estas acumulaciones patológicas de fluido, el desarrollo de estas metodologías es especialmente difícil debido a su naturaleza difusa, sin bordes definidos. En esta tesis doctoral proponemos diferentes metodologías CAD con el objetivo de ayudar a las personas expertas del dominio a detectar y comprender mejor estas patologías. Además, expandimos las metodologías desarrolladas a otras modalidades de imagen médica y afecciones, como al análisis de neovascularizaciones maculares en Angiografía OCT y al diagnóstico de COVID-19 mediante radiografías torácicas., [Resumo] A acumulación de líquido intrarretiniano é tanto o síntoma común como culpable das principais causas de cegueira nos países desenvolvidos: a dexeneración macular asociada á idade e o edema macular diabético. Para o seu diagnóstico, os expertos no campo empregan imaxes de tomografía de coherencia óptica (OCT), que proporcionan representacións transversais non invasivas das estruturas retinianas. Non obstante, ao igual que calquera modalidade de imaxe médica, a OCT vese influenciada por múltiples factores que afectan a s´ua calidade e a súa posterior interpretación. Xunto coa subxectividade dos expertos humanos, estes factores poden afectar significativamente ao proceso diagn´ostico, ao tratamento e á calidade de vida das persoas afectadas (particularmente nestas patoloxías onde unha detección precoz é crucial). Para abordar estes desafíos, desenvólvense metodoloxías de diagnóstico asistido por ordenador (CAD), que ofrecen unha capa de abstracción da información presente nas imaxes. Non obstante, no escenario particular das acumulacións patolóxicas de líquido, o desenvolvemento destas metodoloxías é especialmente difícil debido a súa natureza difusa, sen bordes definidos. Nesta tese de doutoramento propoñemos diferentes metodoloxías de CAD co obxectivo de axudar ás persoas expertas do campo a detectar e comprender mellor estas patoloxías. Ademais, expandimos as metodoloxías desenvoltas a outras modalidades de imaxe médica e patoloxías, como a an´alise de neovascularizacións maculares en Anxiografía OCT e ao diagnóstico da COVID-19 mediante a análise de radiografías torácicas.

[Resumen]: El proyecto ”Pk-Pharm” es una iniciativa innovadora creada en colaboración entre el Grupo de Visión Artificial y Reconocimiento de Patrones (VARPA) de la Universidad de A Coruña (UDC), la Unidad de Investigación e Innovación del Servicio de Farmacia (CHUS) y el Grupo de Farmacología Clínica del IDIS. Su objetivo principal es desarrollar una plataforma de software de última generación que permita la dosificación personalizada de fármacos, adaptada a las características específicas de cada paciente. La disciplina de la farmacocinética se enfoca en estudiar cómo los fármacos interactúan con el cuerpo humano. Debido a que la farmacocinética individual puede variar significativamente por factores genéticos, ambientales y fisiológicos, la dosificación adecuada de medicamentos puede ser un desafío en la práctica clínica. Por esta razón, la posología personalizada busca mejorar la eficacia y seguridad de los tratamientos farmacológicos al adaptar la dosis de cada medicamento a las necesidades y características específicas de cada paciente. La propuesta consiste en desarrollar una plataforma web que incluya un sistema para monitorear y administrar medicamentos de manera individualizada, considerando factores como la edad, el peso, la altura y otros parámetros antropométricos, analíticos, extracorpóreos y farmacogenéticos. Esto permitirá optimizar la eficacia de los tratamientos farmacológicos, minimizar los efectos secundarios y reducir los riesgos asociados a la medicación. Para lograr este objetivo, ”Pk-Pharm” utilizará modelos farmacocinéticos y farmacodinámicos avanzados, que determinarán la dosis óptima de cada medicamento en función de las características individuales del paciente, así como las posibles interacciones con otros medicamentos que esté tomando. La aplicación de estas metodologías permitirá mejorar la precisión en la estimación de los parámetros farmacocinéticos, optimizando así la terapia farmacológica y reduciendo la variabilidad en las respues, [Abstract]: The ”Pk-Pharm” project is an innovative initiative created in collaboration between the Artificial Vision and Pattern Recognition Group (VARPA) of the University of A Coruña (UDC), the Research and Innovation Unit of the Pharmacy Service (CHUS) and the Clinical Pharmacology Group of IDIS. Its main objective is to develop a state-of-the-art software platform that allows the personalised dosage of drugs, adapted to the specific characteristics of each patient. The discipline of pharmacokinetics focuses on the study of how drugs interact with the human body. Because individual pharmacokinetics can vary significantly by genetic, environmental and physiological factors, appropriate drug dosing can be a challenge in clinical practice. For this reason, personalised dosing aims to improve the efficacy and safety of drug treatments by tailoring the dose of each drug to the specific needs and characteristics of each patient. The proposal consists of developing a web platform that includes a system for monitoring and administering drugs in an individualised manner, taking into account factors such as age, weight, height and other anthropometric, analytical, extracorporeal and pharmacogenetic parameters. This will optimise the efficacy of drug treatments, minimise side effects and reduce the risks associated with medication. To achieve this goal, ”Pk-Pharm” will use advanced pharmacokinetic and pharmacodynamic models, which will determine the optimal dose of each drug according to the individual characteristics of the patient, as well as possible interactions with other drugs that the patient is taking. The application of these methodologies will improve the accuracy of pharmacokinetic parameter estimation, thereby optimising drug therapy and reducing variability in patients’ responses to drugs. This platform will be very useful for doctors and pharmacists, as it will increase the accuracy of drug dosing and allow closer and more rigorous monitoring of patient treat

[Resumen]: La enfermedad de Alzheimer representa uno de los desafíos más apremiantes en el ámbito médico y farmacológico actual. Identificar biomarcadores efectivos y desarrollar técnicas de diagnóstico temprano son elementos clave para combatir esta epidemia silenciosa. En este entorno, el presente Trabajo de Fin de Grado (TFG) explora un enfoque novedoso: el uso de imágenes de tomografía de coherencia óptica (OCT) retinianas obtenidas de ratones transgénicos PS19 tau. Estos ratones se caracterizan por su susceptibilidad a desarrollar sintomatología similar a la del alzhéimer en humanos, lo que los convierte en modelos preclínicos ideales para esta investigación. Nuestro enfoque innovador emplea técnicas de aprendizaje profundo para construir un sistema de segmentación automática tridimensional. Este sistema es capaz de extraer biomarcadores retinianos críticos que pueden correlacionarse con el desarrollo de la enfermedad de Alzheimer. Este conjunto de algoritmos forma parte de un pipeline de diagnóstico automatizado, complementado por un modelo de clasificación específico para estos biomarcadores. Los resultados iniciales son alentadores, mostrando un alto grado de precisión y validando la aplicabilidad de nuestra metodología en entornos preclínicos y farmacológicos. Este enfoque no solo mejora la velocidad y la eficacia del diagnóstico, sino que también reduce la subjetividad y la variabilidad asociadas a la interpretación humana. El proyecto se desglosa en dos fases fundamentales: la primera enfocada en la segmentación automática de imágenes OCT para la identificación de biomarcadores relevantes, y la segunda en la clasificación robusta de dichos biomarcadores para un diagnóstico preciso de la enfermedad de Alzheimer., [Abstract]: The Alzheimer’s disease represents one of the most pressing challenges in today’s medical and pharmacological fields. Identifying effective biomarkers and developing early diagnosis techniques are key elements in combating this silent epidemic. Within this context, the present Thesis explores a novel approach: the use of Optical Coherence Tomography (OCT) retinal images obtained from PS19 tau transgenic mice. These mice are characterized by their susceptibility to develop symptoms similar to Alzheimer’s in humans, making them ideal preclinical models for this research. Our innovative approach employs deep learning techniques to build a three-dimensional automatic segmentation system. This system is capable of extracting critical retinal biomarkers that can be correlated with the development of Alzheimer’s disease. This set of algorithms forms part of an automated diagnostic pipeline, complemented by a specific classification model for these biomarkers. The initial results are encouraging, showing a high degree of accuracy and validating the applicability of our methodology in preclinical and pharmacological settings. This approach not only improves the speed and efficacy of the diagnosis but also reduces the subjectivity and variability associated with human interpretation. The project is broken down into two fundamental phases: the first focuses on the automatic segmentation of OCT images for the identification of relevant biomarkers, and the second on the robust classification of these biomarkers for an accurate diagnosis of Alzheimer’s disease.

[Resumen]: Las enfermedades neurodegenerativas afectan a millones de personas en todo el mundo y pueden disminuir significativamente la calidad de vida y en algunos casos llevar a la muerte. Recientes estudios han demostrado que las enfermedades neurodegenerativas no solo afectan el cerebro, sino que también pueden tener un impacto en la retina y el nervio óptico. Segmentar e identificar estas enfermedades a partir de tomografía de coherencia óptica (OCT) es un paso crítico para un diagnóstico temprano. No obstante, el desafío radica en que los efectos de estas enfermedades en la retina suelen ser sutiles y difíciles de distinguir. En este trabajo, presentamos una metodología robusta basada en técnicas de Geometric Deep Learning para proporcionar un apoyo objetivo y mejorar el diagnóstico. La metodología se compone de dos fases. En la primera, se realiza una segmentación automática de las capas de la retina en imágenes OCT empleando nnU-Net, una arquitectura de aprendizaje profundo clásico. En la segunda, utilizamos los resultados de la segmentación en diferentes arquitecturas de Geometric Deep Learning como PointNet++ y DGCNN, aprovechando la información geométrica de todo el volumen OCT. Una de las mayores ventajas de utilizar Geometric Deep Learning en este contexto es su eficiencia en el uso de datos. Estas arquitecturas son especialmente aptas para trabajar con datos estructurados de manera irregular, como son las nubes de puntos en imágenes tridimensionales. Esto nos permite hacer un uso más eficiente de los datos disponibles, extrayendo patrones y relaciones complejas que serían difíciles de capturar con métodos convencionales. Además, la capacidad de estas arquitecturas para trabajar con diferentes números de puntos en una nube ofrece un equilibrio eficaz entre precisión y eficiencia computacional. El potencial de este estudio es considerable. Mejorar la precisión del diagnóstico de enfermedades neurodegenerativas a través de técnicas no invasivas implica un, [Abstract]: Neurodegenerative diseases affect millions of people worldwide, significantly diminishing their quality of life and, in some cases, leading to death. Recent studies have shown that neurodegenerative diseases not only impact the brain but can also affect the retina and optic nerve. Segmenting and identifying these diseases through Optical Coherence Tomography (OCT) is a critical step for early diagnosis. However, the challenge lies in the fact that the effects of these diseases on the retina are often subtle and difficult to distinguish. In this work, we present a robust methodology based on Geometric Deep Learning techniques to provide objective support and improve diagnosis. The methodology consists of two phases. In the first phase, we perform automatic segmentation of the retinal layers in OCT images using nnU-Net, a classic deep learning architecture. In the second phase, we employ the segmentation results in different Geometric Deep Learning architectures like PointNet++ and DGCNN, leveraging the geometric information across the entire OCT volume. One of the significant advantages of using Geometric Deep Learning in this context is its data efficiency. These architectures are particularly suited for working with irregularly structured data, such as point clouds in three-dimensional images. This allows us to make more efficient use of available data, extracting complex patterns and relationships that would be difficult to capture with conventional methods. Additionally, the capability of these architectures to work with different numbers of points in a cloud offers an effective balance between accuracy and computational efficiency. The potential of this study is considerable. Improving the accuracy of diagnosing neurodegenerative diseases through non-invasive techniques implies a safer, earlier, and more accurate diagnosis, which can lead to more effective treatments and an improved quality of life for the patient.

[Abstract] Computer-Aided Screening (CAS) systems are getting popularity in disease diagnosis. Modern CAS systems exploit data driven machine learning algorithms including supervised and unsupervised methods. In medical imaging, annotating pathological samples are much harder and time consuming work than healthy samples. Therefore, there is always an abundance of healthy samples and scarcity of annotated and labelled pathological samples. Unsupervised anomaly detection algorithms can be implemented for the development of CAS system using the largely available healthy samples, especially when disease/nodisease decision is important for screening. This thesis proposes unsupervised machine learning methodologies for anomaly detection in retinal fundus images. A novel patchbased image reconstructor architecture for DR detection is presented, that addresses the shortcomings of standard autoencoders-based reconstructors. Furthermore, a full-size image based anomaly map generation methodology is presented, where the potential DR lesions can be visualized at the pixel-level. Afterwards, a novel methodology is proposed to extend the patch-based architecture to a fully-convolutional architecture for one-shot full-size image reconstruction. Finally, a novel methodology for supervised DR classification is proposed that utilizes the anomaly maps.

[Resumen]: La optimización del uso de áreas de trabajo, recursos y tiempo es fundamental para mejorar la eficiencia, reducir costes y aumentar la productividad en cualquier organización. En entornos industriales con elevada generación de residuos, la monitorización del uso de contenedores en cuanto a capacidad disponible, nivel de ocupación y necesidad de vaciado es crucial para asegurar la continuidad de todas las fases de las cadenas productivas. Con el auge de la industria 4.0 y los avances en métodos computacionales basados en aprendizaje automático para el procesado automático de imagen y vídeo, cada vez son más las organizaciones que demandan la automatización de procesos y tareas para lograr una gestión global e inteligente de su infraestructura y operaciones. En este trabajo de fin de grado se propone el desarollo de un método para el control automatizado del uso de contenedores industriales a partir de imágenes obtenidas por un sistema de cámaras Closed-Circuit Television (CCTV) ubicado en un entorno industrial real. El método propuesto consta de una primera fase para la segmentación de las zonas de interés en la imagen, como las regiones ocupadas por residuos y una segunda fase para la estimación del nivel de ocupación de cada contenedor individual y su clasificación en base a si requiere ser vaciado o no. Para abordar estas tareas se realiza un análisis de aproximaciones basadas en técnicas clásicas de machine learning y aproximaciones basadas en deep learning. Se consideran diferentes configuraciones de parámetros y se realiza un estudio comparativo para seleccionar la aproximación más adecuada para este domino. Adicionalmente, la metodología desarrollada ha sido integrada en un servicio de videovigilancia de una plataforma para la gestión global e inteligente de infraestructuras industriales, proporcionando una herramienta de apoyo que permite la monitorización fiable y objetiva del uso de contenedores y la generación y envío de las alertas correspondie, [Abstract]: Optimizing the use of work areas, resources, and time is essential for improving efficiency, reducing costs, and increasing productivity in any organization. In industrial environments with high waste generation, monitoring the use of containers in terms of available capacity, occupancy level, and the need for emptying is crucial to ensure the continuity of all phases of the production chains. With the rise of Industry 4.0 and advancements in computational methods based on machine learning for automated image and video processing, more and more organizations are demanding process and task automation to achieve comprehensive and intelligent management of their infrastructure and operations. This thesis proposes developing a method for the automated control of industrial container usage based on images obtained from a Closed-Circuit Television (CCTV) camera system located in a real industrial environment. The proposed method consists of a first phase for segmenting the areas of interest in the image, such as regions occupied by waste, and a second phase for estimating the occupancy level of each individual container and classifying it based on whether it requires emptying or not. To address these tasks, an analysis of approaches based on classical machine learning techniques and deep learning approaches is performed. Different parameter configurations are considered, and a comparative study is conducted to select the most suitable approach for this domain. Additionally, the developed methodology has been integrated into a video surveillance service of a platform for comprehensive and intelligent management of industrial infrastructures, providing a supportive tool that allows reliable and objective monitoring of container usage and the generation and sending of corresponding alerts.

[Resumen]: La degeneración macular asociada a la edad (DMAE) es una de las principales causas de pérdida de visión en individuos mayores de 50 años. Los primeros indicadores de esta enfermedad, las drusas, son minúsculas acumulaciones de material que se forman en la retina. Detectar, segmentar y cuantificar estas drusas en las imágenes de tomografía de coherencia óptica (OCT) es un paso crítico para el diagnóstico temprano y seguimiento de la DMAE. Sin embargo, este proceso puede ser un desafío debido a las variaciones en el tamaño, la forma y la distribución de las drusas. En este trabajo, presentamos una metodología robusta basada en técnicas de aprendizaje profundo y procesamiento de imágenes para mejorar la precisión y la eficiencia en la segmentación, identificación y cuantificación de drusas. La metodología propuesta consta de dos fases. En la primera, se realiza una segmentación automática de las drusas utilizando diferentes arquitecturas de aprendizaje profundo, aprovechando su capacidad para manejar las variaciones en las imágenes OCT. En la segunda fase, implementamos un módulo para la identificación y cuantificación automática de las drusas, proporcionando información detallada sobre su número y distribución en la retina. El impacto potencial de este estudio es considerable. La mejora en la precisión de la segmentación de drusas abre la puerta a diagnósticos más tempranos y precisos de la DMAE, lo que puede llevar a tratamientos más efectivos y resultados mejorados para los pacientes. Además, al abordar un desafío clave en la oftalmología con técnicas de inteligencia artificial y procesamiento de imágenes, este trabajo podría sentar las bases para futuras investigaciones en la detección automática de otras características oculares y patologías., [Abstract]: Age-related macular degeneration (AMD) is a leading cause of vision loss in individuals over the age of 50. Early indicators of this disease, drusen, are tiny accumulations of material that form in the retina. Detecting, segmenting, and quantifying these drusen in Optical Coherence Tomography (OCT) images is a critical step in early AMD diagnosis and monitoring. However, this process can be challenging due to variations in the size, shape, and distribution of drusen. In this study, we introduce a robust methodology based on deep learning techniques and image processing to improve precision and efficiency in drusen segmentation, identification, and quantification. The proposed methodology consists of two phases. In the first, automatic segmentation of drusen is performed using different deep learning architectures, leveraging their ability to handle variations in OCT images. In the second phase, we implement a module for automatic drusen identification and quantification, providing detailed information about their number and distribution in the retina. The potential impact of this study is significant. The improvement in drusen segmentation accuracy paves the way for earlier and more accurate AMD diagnoses, which could lead to more effective treatments and improved outcomes for patients. Moreover, by addressing a key challenge in ophthalmology with artificial intelligence and image processing techniques, this work could lay the groundwork for future research into the automatic detection of other ocular features and pathologies.

[Abstract]: Optical coherence tomography angiography (OCTA) is a non-invasive ophthalmic imaging modality that is widely used in clinical practice. Recent technological advances in OCTA allow imaging of blood flow deeper than the retinal layers, at the level of the choriocapillaris (CC), where a granular image is obtained showing a pattern of bright areas, representing blood flow, and a pattern of small dark regions, called flow voids (FVs). Several clinical studies have reported a close correlation between abnormal FVs distribution and multiple diseases, so quantifying changes in FVs distribution in CC has become an area of interest for many clinicians. However, CC OCTA images present very complex features that make it difficult to correctly compare FVs during the monitoring of a patient. In this work, we propose fully automatic approaches for the segmentation and monitoring of FVs in CC OCTA images. First, a baseline approach, in which a fully automatic segmentation methodology based on local contrast enhancement and global thresholding is proposed to segment FVs and measure changes in their distribution in a straightforward manner. Second, a robust approach in which, prior to the use of our segmentation methodology, an unsupervised trained neural network is used to perform a deformable registration that aligns inconsistencies between images acquired at different time instants. The proposed approaches were tested with CC OCTA images collected during a clinical study on the response to photodynamic therapy in patients affected by chronic central serous chorioretinopathy (CSC), demonstrating their clinical utility. The results showed that both approaches are accurate and robust, surpassing the state of the art, therefore improving the efficacy of FVs as a biomarker to monitor the patient treatments. This gives great potential for the clinical use of our methods, with the possibility of extending their use to other pathologies or treatments associated with this type

[Abstract]: During the last years, deep learning techniques have emerged as powerful alternatives to solve biomedical image analysis problems. However, the training of deep neural networks usually needs great amounts of labeled data to be done effectively. This is even more critical in the case of biomedical imaging due to the added difficulty of obtaining data labeled by experienced clinicians. To mitigate the impact of data scarcity, one of the most commonly used strategies is transfer learning. Nevertheless, the success of this approach depends on the effectiveness of the available pre-training techniques for learning from little or no labeled data. In this work, we explore the application of the Context Encoder paradigm for transfer learning in the domain of retinal image analysis. To this aim, we propose several approaches that allow to work with full resolution images and improve the recognition of the retinal structures. In order to validate the proposals, the Context Encoder pre-trained models are fine-tuned to perform two relevant tasks in the domain: vessels segmentation and fovea localization. The experiments performed on different public datasets demonstrate that the proposed Context Encoder approaches allow mitigating the impact of data scarcity, being superior to previous alternatives in this domain.

[Abstract]: Background and Objectives: Age-related macular degeneration (AMD) is a degenerative disorder affecting the macula, a key area of the retina for visual acuity. Nowadays, AMD is the most frequent cause of blindness in developed countries. Although some promising treatments have been proposed that effectively slow down its development, their effectiveness significantly diminishes in the advanced stages. This emphasizes the importance of large-scale screening programs for early detection. Nevertheless, implementing such programs for a disease like AMD is usually unfeasible, since the population at risk is large and the diagnosis is challenging. For the characterization of the disease, clinicians have to identify and localize certain retinal lesions. All this motivates the development of automatic diagnostic methods. In this sense, several works have achieved highly positive results for AMD detection using convolutional neural networks (CNNs). However, none of them incorporates explainability mechanisms linking the diagnosis to its related lesions to help clinicians to better understand the decisions of the models. This is specially relevant, since the absence of such mechanisms limits the application of automatic methods in the clinical practice. In that regard, we propose an explainable deep learning approach for the diagnosis of AMD via the joint identification of its associated retinal lesions. Methods: In our proposal, a CNN with a custom architectural setting is trained end-to-end for the joint identification of AMD and its associated retinal lesions. With the proposed setting, the lesion identification is directly derived from independent lesion activation maps; then, the diagnosis is obtained from the identified lesions. The training is performed end-to-end using image-level labels. Thus, lesion-specific activation maps are learned in a weakly-supervised manner. The provided lesion information is of high clinical interest, as it allows clinicians to a

[Abstract]: Diabetes represents one of the main causes of blindness in developed countries, caused by fluid accumulations in the retinal layers. The clinical literature defines the different types of diabetic macular edema (DME) as cystoid macular edema (CME), diffuse retinal thickening (DRT), and serous retinal detachment (SRD), each with its own clinical relevance. These fluid accumulations do not present defined borders that facilitate segmentational approaches (specially the DRT type, usually not taken into account by the state of the art for this reason) so a diffuse paradigm is used for its detection and visualization. In this paper, we propose three novel approaches for the representation and characterization of these types of DME. A baseline proposal, using a convolutional neural network as backbone, another based on transfer learning from a general domain, and a third approach exploiting information of regions without a defined label. Overall, our baseline proposal obtained an AUC of 0.9583 ± 0.0093, the approach pretrained with a general-domain dataset an AUC of 0.9603 ± 0.0087, and the approach pretrained in the domain taking advantage of uncertainty, an AUC of 0.9619 ± 0.0073.

[Abstract]: In optical coherence tomography (OCT), there is a trade-off between the scanning time and image quality, leading to a scarcity of high quality data. OCT platforms provide different scanning presets, producing visually distinct images, limiting their compatibility. In this work, a fully automatic methodology for the unpaired visual conversion of the two most prevalent scanning presets is proposed. Using contrastive unpaired translation generative adversarial architectures, low quality images acquired with the faster Macular Cube preset can be converted to the visual style of high visibility Seven Lines scans and vice-versa. This modifies the visual appearance of the OCT images generated by each preset while preserving natural tissue structure. The quality of original and synthetic generated images was compared using BRISQUE. The synthetic generated images achieved very similar scores to original images of their target preset. The generative models were validated in automatic and expert separability tests. These models demonstrated they were able to replicate the genuine look of the original images. This methodology has the potential to create multi-preset datasets with which to train robust computer-aided diagnosis systems by exposing them to the visual features of different presets they may encounter in real clinical scenarios without having to obtain additional data.

[Abstract]: One of the main issues with deep learning is the need of a significant number of samples. We intend to address this problem in the field of Optical Coherence Tomography (OCT), specifically in the context of Diabetic Macular Edema (DME). This pathology represents one of the main causes of blindness in developed countries and, due to the capturing difficulties and saturation of health services, the task of creating computer-aided diagnosis (CAD) systems is an arduous task. For this reason, we propose a solution to generate samples. Our strategy employs image-to-image Generative Adversarial Networks (GAN) to translate a binary mask into a realistic OCT image. Moreover, thanks to the clinical relationship between the retinal shape and the presence of DME fluid, we can generate both pathological and non-pathological samples by altering the binary mask morphology. To demonstrate the capabilities of our proposal, we test it against two classification strategies of the state-of-the-art. In the first one, we evaluate a system fully trained with generated images, obtaining 94.83% accuracy with respect to the state-of-the-art. In the second case, we tested it against a state-of-the-art expert model based on deep features, in which it also achieved successful results with a 98.23% of the accuracy of the original work. This way, our methodology proved to be useful in scenarios where data is scarce, and could be easily adapted to other imaging modalities and pathologies where key shape constraints in the image provide enough information to recreate realistic samples.

[Resumen]: La enfermedad de Alzheimer (EA) es una enfermedad neurodegenerativa caracterizada por la acumulación de placas de ß-amiloide (ßA) y proteína TAU que se depositan en las placas seniles y causan efectos tóxicos directos sobre las neuronas. Los pacientes presentan alteraciones visuales y cambios morfológicos identificables en la retina mediante el uso de la tomografía de coherencia óptica (OCT). El objetivo es caracterizar los cambios en la morfología retiniana mediante un OCT preclínico en un modelo animal de EA con tauopatía.

[Abstract]: X-ray analysis of the lungs was the main method to assess the degree of affliction of SARS-COV-2. Due to the high contagiousness of this pathology, this assessment was conducted using portable X-ray devices. Automatic methodologies were proposed to compensate the image quality of said portable X-ray devices. However, these methodologies were shown to be exploiting external information (such as pacemakers or ventilators present in the images) to determine the severity. For this reason, we present a methodology specially designed to reduce the effect on an automatic methodology of these extraneous artifacts. We extract the lung region and we perform a screening of the presence of the pathology using only the pulmonary region. Finally, to ascertain the performance of the system (and provide explainability to the clinical experts), we generate the corresponding activation maps. The presented methodology has achieved a more than satisfactory performance in all the scenarios and the activation maps clearly indicate that the system is successfully using information from the lung region while excluding elements unrelated to the disease.

[Abstract]: COVID-19 is a disease whose gold standard diagnosis tool, RT-PCR, is unable to provide accurate quantification of its severity in a given patient. Currently, this assessment can be performed with the help of chest X-ray imaging visualization that, however, is a manual, tedious and time-consuming task. In this context, Computer-Aided Diagnosis (CAD) systems are very useful to facilitate the work of clinical specialists in these complex diagnostic tasks, especially in view of recent advances in deep learning techniques in the field of medical image analysis. Despite their great potential, deep learning strategies require a large amount of labelled data, which is often scarce in the context of COVID-19 pandemic. To mitigate these problems, in this work we propose the use of a image translation paradigm, the Cycle-Consistent Adversarial Networks (CycleGAN) to generate a novel set of synthetic images with the aim to improve an automatic COVID-19 screening system using portable chest X-ray images.

El Alzhéimer (AD), el temblor esencial (ET), la esclerósis múltiple (MS) o el Párkinson (PD) son enfermedades neurodegenerativas (END) que están correlacionadas con cambios en algunas capas retinales clave. Las tomografías de coherencia óptica (OCT) pueden proporcionar información detallada desde diferentes perspectivas para analizar esas capas. Proponemos un sistema automático que: Segmenta las capas RNFL y GCL-BM de vólumenes 3D e imágenes OCT retinales de mácula y disco óptico. Extrae biomarcadores computacionales como deep features o grosor. Hace un cribado combinando información de ambas vistas OCT.

[Abstract]: Optical Coherence Tomography (OCT) is the major diagnostic tool for the leading cause of blindness in developed countries: Diabetic Macular Edema (DME). Depending on the type of fluid accumulations, different treatments are needed. In particular, Cystoid Macular Edemas (CMEs) represent the most severe scenario, while Diffuse Retinal Thickening (DRT) is an early indicator of the disease but a challenging scenario to detect. While methodologies exist, their explanatory power is limited to the input sample itself. However, due to the complexity of these accumulations, this may not be enough for a clinician to assess the validity of the classification. Thus, in this work, we propose a novel approach based on multi-prototype networks with vision transformers to obtain an example-based explainable classification. Our proposal achieved robust results in two representative OCT devices, with a mean accuracy of 0.9099 ± 0.0083 and 0.8582 ± 0.0126 for CME and DRT-type fluid accumulations, respectively.

[Absctract]: The Epiretinal Membrane (ERM) is an ocular pathology that causes visual distortion. In order to detect and treat the ERM, ophthalmologists visually inspect Optical Coherence Tomography (OCT) images.This is a costly and subjective process. In this work, we present three different fully automatic, end-to-end approaches that make use of multi-task learning to simultaneously screen for and segment ERM symptoms in OCT images. These approaches were implemented into three architectures that capitalise on the way the models share a single architecture for the two complementary tasks.

[Absctract]: The Epiretinal Membrane (ERM) is an ocular pathology that can cause permanent visual loss if left untreated for long. Despite its transparency, it is possible to visualise the ERM in Optical Coherence Tomography (OCT) images. In this work, we present a study on the impact of the analysis region on the performance of an automatic ERM segmentation methodology using OCT images. For this purpose, we tested 5 different sliding windows sizes ranging from to pixels to calibrate the impact of the field of view under analysis. Furthermore, 3 different approaches are proposed to enable the analysis of the regions close to the edges of the images. The proposed approaches provided satisfactory results, with each of them interacting differently with the variations in window size.

[Abstract]: COVID-19 is a challenging disease that was declared as global pandemic in March 2020. As the main impact of this disease is located in the pulmonary regions, chest X-ray devices are very useful to understand the severity of the disease on each patient. In order to reduce the risk of cross-contamination, the radiologists are recommended to use portable devices instead of fixed machinery, as these devices are easier to decontaminate. Moreover, the development of reliable and robust methodologies of computer-aided diagnosis systems is very relevant to reduce the workload that expert clinicians are experiencing in the current moment. In this work, we propose a comprehensive analysis of the deep features extracted from portable chest X-ray captures to perform a COVID-19 screening. We also study the optimal characterization of the problem with a lower dimensionality, contrasting the results of the feature selection methods that were chosen. Results demonstrated that the proposed approach is robust and reliable, obtaining a 90.43% of accuracy for the test set, using only 46.85% of the deep features in the context of poor quality and low detail X-ray images obtained from portable devices.

[Absctract]: Retinal Optical Coherence Tomography (OCT) allows the non-invasive direct observation of the central nervous system, enabling the measurement and extraction of biomarkers from neural tissue that can be helpful in the assessment of ocular, systemic and Neurological Disorders (ND). Deep learning models can be trained to segment the retinal layers for biomarker extraction. However, the onset of ND can have an impact on the neural tissue, which can lead to the degraded performance of models not exposed to images displaying signs of disease during training. We present a fully automatic approach for the retinal layer segmentation in multiple neurodegenerative disorder scenarios, using an annotated dataset of patients of the most prevalent NDs: Alzheimer's disease, Parkinson's disease, multiple sclerosis and essential tremor, along with healthy control patients. Furthermore, we present a two-part, comprehensive study on the effects of ND on the performance of these models. The results show that images of healthy patients may not be sufficient for the robust training of automated segmentation models intended for the analysis of ND patients, and that using images representative of different NDs can increase the model performance. These results indicate that the presence or absence of patients of ND in datasets should be taken into account when training deep learning models for retinal layer segmentation, and that the proposed approach can provide a valuable tool for the robust and reliable diagnosis in multiple scenarios of ND.

[Abstract]: Optical Coherence Tomography Angiography or OCTA represents one of the main means of diagnosis of Age-related Macular Degeneration (AMD), the leading cause of blindness in developed countries. This eye disease is characterized by Macular Neovascularization (MNV), the formation of vessels that tear through the retinal tissues. Four types of MNV can be distinguished, each representing different levels of severity. Both the aggressiveness of the treatment and the recovery of the patient rely on an early detection and correct diagnosis of the stage of the disease. In this work, we propose the first fully-automatic grading methodology that considers all the four clinical types of MNV at the three most relevant OCTA scanning depths for the diagnosis of AMD. We perform both a comprehensive ablation study on the contribution of said depths and an analysis of the attention maps of the network in collaboration with experts of the domain. Our proposal aims to ease the diagnosis burden and decrease the influence of subjectivity on it, offering a explainable grading through the visualization of the attention of the expert models. Our grading proposal achieved satisfactory results with an AUC of 0.9224 ± 0.0381. Additionally, the qualitative analysis performed in collaboration with experts revealed the relevance of the avascular plexus in the grading of all three types of MNV (despite not being directly involved in some of them). Thus, our proposal is not only able to robustly detect MNV in complex scenarios, but also aided to discover previously unconsidered relationships between plexuses.

[Abstract]: La COVID-19 es una enfermedad pulmonar infecciosa causante de la pandemia mundial del año 2020. En los momentos críticos de las emergencias sanitarias, el equipo médico debe tomar decisiones importantes en un contexto de recursos limitados. En este contexto, los métodos de diagnóstico asistido por ordenador pueden ayudar en esa toma de decisiones, permitiendo identificar a los pacientes de alto riesgo. Esto se puede realizar utilizando información extraída de historiales clínicos electrónicos, tests de laboratorio o estudios de imagen. En este trabajo, presentamos un nuevo método eficiente y totalmente automático para estimar el riesgo de pacientes COVID-19, utilizando la fusión de datos multimodales con características clínicas y características profundas extraídas a partir de imágenes de radiografía torácica. Esta estimación se efectúa en 2 escenarios críticos: riesgo de hospitalización y riesgo de fallecimiento. Los resultados demuestran un gran desempeño por parte de los clasificadores, estimando ambos riesgos con una gran precisión y utilizando únicamente un subconjunto muy reducido de un conjunto de características originales notablemente más extenso. Esta reducción de la dimensionalidad en el conjunto de datos es muy ventajosa en escenarios donde los recursos computacionales son limitados. Este método totalmente automático presenta un potencial prometedor para mejorar el proceso de toma de decisiones clínicas y una mejor gestión de los recursos médicos, no solo en el contexto de la COVID-19 sino también en otros escenarios clínicos.

[Abstract]: Patients at high risk of contracting COVID-19 require specialized monitoring throughout their illness to ensure optimal treatment at each stage. To support this monitoring, Computer-Aided Diagnosis (CAD) methods analyze clinical data to estimate the most likely outcome for each patient, using various clinical variables such as symptoms, medical history, and laboratory results to predict outcomes. Despite the numerous proposals for COVID-19 diagnosis using CAD methods, the lack of explainability in many machine learning models poses a challenge in incorporating these methods into clinical practice. Additionally, other crucial tasks such as estimating the risk of death or severe forms of the disease must be considered to identify cases that require greater monitoring. To overcome these challenges, we propose an explainable methodology for estimating the risk of hospitalization and death in COVID-19 patients using clinical data. Our methodology employs four machine learning algorithms, three feature selection methods, and a decision tree to provide explainability. Our approach achieves an accuracy of 86.16% ± 0.74% for the estimation of hospitalization risk with 29 features, and an accuracy of 86.40% ± 1.80% for the estimation of the risk of death with 26 features. Moreover, our methodology provides valuable insights into the relationship between clinical variables and patient outcomes, which can inform more robust and informed clinical decision-making and improve our understanding of the disease. We demonstrate the potential of our transparent and effective CAD methods to support clinical decision-making in COVID-19 patient care and further research, offering a promising solution to overcome the challenges in incorporating CAD methods into clinical practice.

Purpose : The retina is the most accessible part of the central nervous system, allowing its non-invasive exploration and measurement. Optical Coherence Tomography (OCT) offers an objective monitoring method of progression in Neurodegenerative Disease (NDD), enabling the extraction of biomarkers such as retinal layer thickness. Machine learning models allow the automatic and repeatable measurement of the retinal layers and enable an early diagnosis of NDD. These need to be trained on annotated images representative of the visual patterns that characterise these diseases. We present a study in the automatic measurement of retinal layer thickness in patients of different NDDs and an assessment of the mutual compatibility of models trained in representative images of these diseases. Methods : Five independent samples of multiple sclerosis, Alzheimer's disease, Parkinson's disease and essential tremor patients, along with healthy controls were prospectively recruited (N=50). Macula centred OCT volumes from these patients were annotated with the area of the Retinal Nerve Fibre Layer (RNFL) and between the Ganglion Cell Layer and Bruch's Membrane (GCL-BM), for 1250 B-scans in total. In a first experiment, a series of deep learning models were trained on every NDD but one and evaluated in terms of their ability to segment the retinal layers of the unseen NDD. In a second experiment, the models were trained on images from a specific NDD and then evaluated on each of the other ones. Results : The average thickness for each layer was measured and separately compared for each NDD using a one-way ANOVA test. This test found no significant differences between the thickness of either RNFL or GCL-BM (p>0.05). The results show that the models are able to accurately segment the retinal layers, with an overall Dice coefficient of 0.96±0.02 for Experiment 1 and 0.95±0.03 for Experiment 2. However, these results do not translate equally for every NDD. The models trained in diseases suc

[Resumen] Se presenta un innovador método de deep learning para la segmentación automatizada en 3D de las regiones de fluido en imágenes de Tomografía de Coherencia ´Óptica (OCT) de pacientes con coriorretinopatía serosa central (CSCR), seguida de un análisis de respuesta a la Terapia Fotodinámica (PDT) en pacientes con CSCR. Este método no solo reduce sustancialmente el tiempo y esfuerzo requeridos para la segmentación, sino que también ofrece una técnica estandarizada, fomentando estudios de investigación a gran escala. Para llevar a cabo el trabajo utilizamos un conjunto de datos robusto compuesto por 2,769 imágenes OCT, logrando resultados altamente satisfactorios que superan a las demás propuestas del estado del arte. Esta investigación impulsa la integración del deep learning en la práctica clínica, mejorando la gestión de la CSCR al permitir la formulación de estrategias de tratamiento personalizadas y una atención optimizada al paciente.

[Absctract]: Background: Glaucoma is the leading global cause of irreversible blindness. Glaucoma patients experience a progressive deterioration of the retinal nervous tissues that begins with a loss of peripheral vision. An early diagnosis is essential in order to prevent blindness. Ophthalmologists measure the deterioration caused by this disease by assessing the retinal layers in different regions of the eye, using different optical coherence tomography (OCT) scanning patterns to extract images, generating different views from multiple parts of the retina. These images are used to measure the thickness of the retinal layers in different regions. Methods: We present two approaches for the multi-region segmentation of the retinal layers in OCT images of glaucoma patients. These approaches can extract the relevant anatomical structures for glaucoma assessment from three different OCT scan patterns: circumpapillary circle scans, macular cube scans and optic disc (OD) radial scans. By employing transfer learning to take advantage of the visual patterns present in a related domain, these approaches use state-of-the-art segmentation modules to achieve a robust, fully automatic segmentation of the retinal layers. The first approach exploits inter-view similarities by using a single module to segment all of the scan patterns, considering them as a single domain. The second approach uses view-specific modules for the segmentation of each scan pattern, automatically detecting the suitable module to analyse each image. Results: The proposed approaches produced satisfactory results with the first approach achieving a dice coefficient of 0.85±0.06 and the second one 0.87±0.08 for all segmented layers. The first approach produced the best results for the radial scans. Concurrently, the view-specific second approach achieved the best results for the better represented circle and cube scan patterns. Conclusions: To the extent of our knowledge, this is the first proposal in the l

[Absctract]: In this work, we analysed 11 imbalance scenarios with female and male COVID-19 patients present in different proportions for the sex analysis, and 6 scenarios where only one specific age range was used for training for the age factor. In each study, 3 different approaches for automatic COVID-19 screening were used: (I) Normal vs COVID-19, (II) Pneumonia vs COVID-19 and (III) Non-COVID-19 vs COVID-19. The present study was validated using two representative public chest X-ray datasets, allowing a reliable analysis to support the clinical decision-making process. The results for the sex-related analysis indicate this factor slightly affects the COVID- 19 deep learning-based systems, although the identified differences are not relevant enough to considerably worsen the system. Regarding the age-related analysis, this factor was observed to be influencing the system in a more consistent way than the sex factor, as it was present in all considered scenarios.