Search

Abstract Long COVID is a multi-systemic disease characterized by the persistence or occurrence of many symptoms that in many cases affect the pulmonary system. These, in turn, may deteriorate the patient’s quality of life making it easier to develop severe complications. Being able to predict this syndrome is therefore important as this enables early treatment. In this work, we investigated three machine learning approaches that use clinical data collected at the time of hospitalization to this goal. The first works with all the descriptors feeding a traditional shallow learner, the second exploits the benefits of an ensemble of classifiers, and the third is driven by the intrinsic multimodality of the data so that different models learn complementary information. The experiments on a new cohort of data from 152 patients show that it is possible to predict pulmonary Long Covid sequelae with an accuracy of up to $$94\%$$ 94 % . As a further contribution, this work also publicly discloses the related data repository to foster research in this field.

Background: With rectum-sparing protocols becoming more common for rectal cancer treatment, this study aimed to predict the pathological complete response (pCR) to preoperative chemoradiotherapy (pCRT) in rectal cancer patients using pre-treatment MRI and a radiomics-based machine learning approach. Methods: We divided MRI-data from 102 patients into a training cohort (n = 72) and a validation cohort (n = 30). In the training cohort, 52 patients were classified as non-responders and 20 as pCR based on histological results from total mesorectal excision. Results: We trained various machine learning models using radiomic features to capture disease heterogeneity between responders and non-responders. The best-performing model achieved a receiver operating characteristic area under the curve (ROC-AUC) of 73% and an accuracy of 70%, with a sensitivity of 78% and a positive predictive value (PPV) of 80%. In the validation cohort, the model showed a sensitivity of 81%, specificity of 75%, and accuracy of 80%. Conclusions: These results highlight the potential of radiomics and machine learning in predicting treatment response and support the integration of advanced imaging and computational methods for personalized rectal cancer management.

Summary: Background: Atypical cartilaginous tumour (ACT) and high-grade chondrosarcoma (CS) of long bones are respectively managed with active surveillance or curettage and wide resection. Our aim was to determine diagnostic performance of X-rays radiomics-based machine learning for classification of ACT and high-grade CS of long bones. Methods: This retrospective, IRB-approved study included 150 patients with surgically treated and histology-proven lesions at two tertiary bone sarcoma centres. At centre 1, the dataset was split into training (n = 71 ACT, n = 24 high-grade CS) and internal test (n = 19 ACT, n = 6 high-grade CS) cohorts, respectively, based on the date of surgery. At centre 2, the dataset constituted the external test cohort (n = 12 ACT, n = 18 high-grade CS). Manual segmentation was performed on frontal view X-rays, using MRI or CT for preliminary identification of lesion margins. After image pre-processing, radiomic features were extracted. Dimensionality reduction included stability, coefficient of variation, and mutual information analyses. In the training cohort, after class balancing, a machine learning classifier (Support Vector Machine) was automatically tuned using nested 10-fold cross-validation. Then, it was tested on both the test cohorts and compared to two musculoskeletal radiologists' performance using McNemar's test. Findings: Five radiomic features (3 morphology, 2 texture) passed dimensionality reduction. After tuning on the training cohort (AUC = 0.75), the classifier had 80%, 83%, 79% and 80%, 89%, 67% accuracy, sensitivity, and specificity in the internal (temporally independent) and external (geographically independent) test cohorts, respectively, with no difference compared to the radiologists (p ≥ 0.617). Interpretation: X-rays radiomics-based machine learning accurately differentiates between ACT and high-grade CS of long bones. Funding: AIRC Investigator Grant.

While extensive research has documented the cognitive changes associated with Parkinson’s disease (PD), a relatively small portion of the empirical literature investigated the language abilities of individuals with PD. Recently, artificial intelligence applied to linguistic data has shown promising results in predicting the clinical diagnosis of neurodegenerative disorders, but a deeper investigation of the current literature available on PD is lacking. This systematic review investigates the nature of language disorders in PD by assessing the contribution of machine learning (ML) to the classification of patients with PD. A total of 10 studies published between 2016 and 2023 were included in this review. Tasks used to elicit language were mainly structured or unstructured narrative discourse. Transcriptions were mostly analyzed using Natural Language Processing (NLP) techniques. The classification accuracy (%) ranged from 43 to 94, sensitivity (%) ranged from 8 to 95, specificity (%) ranged from 3 to 100, AUC (%) ranged from 32 to 97. The most frequent optimal linguistic measures were lexico-semantic (40%), followed by NLP-extracted features (26%) and morphological consistency features (20%). Artificial intelligence applied to linguistic markers provides valuable insights into PD. However, analyzing measures derived from narrative discourse can be time-consuming, and utilizing ML requires specialized expertise. Moving forward, it is important to focus on facilitating the integration of both narrative discourse analysis and artificial intelligence into clinical practice.

Neoadjuvant chemotherapy plus radical surgery could be a safe alternative to chemo-radiation in cervical cancer patients who are not willing to receive radiotherapy. The response to neoadjuvant chemotherapy is the main factor influencing the need for adjunctive treatments and survival. In the present paper we aim to develop a machine learning model based on cervix magnetic resonance imaging (MRI) images to stratify the single-subject risk of cervical cancer. We collected MRI images from 72 subjects. Among these subjects, 28 patients (38.9%) belonged to the “Not completely responding” class and 44 patients (61.1%) belonged to the ’Completely responding‘ class according to their response to treatment. This image set was used for the training and cross-validation of different machine learning models. A robust radiomic approach was applied, under the hypothesis that the radiomic features could be able to capture the disease heterogeneity among the two groups. Three models consisting of three ensembles of machine learning classifiers (random forests, support vector machines, and k-nearest neighbor classifiers) were developed for the binary classification task of interest (“Not completely responding” vs. “Completely responding”), based on supervised learning, using response to treatment as the reference standard. The best model showed an ROC-AUC (%) of 83 (majority vote), 82.3 (mean) [79.9–84.6], an accuracy (%) of 74, 74.1 [72.1–76.1], a sensitivity (%) of 71, 73.8 [68.7–78.9], and a specificity (%) of 75, 74.2 [71–77.5]. In conclusion, our preliminary data support the adoption of a radiomic-based approach to predict the response to neoadjuvant chemotherapy.

The present study was conducted to investigate the potential of radiomics to develop an explainable AI-based system to be applied to ultra-widefield fundus retinographies (UWF-FRTs) with the objective of predicting the presence of the early signs of Age-related Macular Degeneration (AMD) and stratifying subjects with low- versus high-risk of AMD. The ultimate aim was to provide clinicians with an automatic classifier and a signature of objective quantitative image biomarkers of AMD. The use of Machine Learning (ML) and radiomics was based on intensity and texture analysis in the macular region, detected by a Deep Learning (DL)-based macular detector. Two-hundred and twenty six UWF-FRTs were retrospectively collected from two centres and manually annotated to train and test the algorithms. Notably, the combination of the ML-based radiomics model and the DL-based macular detector reported 93% sensitivity and 74% specificity when applied to the data of the centre used for external testing, capturing explainable features associated with drusen or pigmentary abnormalities. In comparison to the human operator’s annotations, the system yielded a 0.79 Cohen κ, demonstrating substantial concordance. To our knowledge, these results are the first provided by a radiomic approach for AMD supporting the suitability of an explainable feature extraction method combined with ML for UWF-FRT.

Abstract Background To evaluate the performance of a decision support system (DSS) based on radiomics and machine learning in predicting the risk of malignancy of ovarian masses (OMs) from transvaginal ultrasonography (TUS) and serum CA-125. Methods A total of 274 consecutive patients who underwent TUS (by different examiners and with different ultrasound machines) and surgery, with suspicious OMs and known CA-125 serum level were used to train and test a DSS. The DSS was used to predict the risk of malignancy of these masses (very low versus medium-high risk), based on the US appearance (solid, liquid, or mixed) and radiomic features (morphometry and regional texture features) within the masses, on the shadow presence (yes/no), and on the level of serum CA-125. Reproducibility of results among the examiners, and performance accuracy, sensitivity, specificity, and area under the curve were tested in a real-world clinical setting. Results The DSS showed a mean 88% accuracy, 99% sensitivity, and 77% specificity for the 239 patients used for training, cross-validation, and testing, and a mean 91% accuracy, 100% sensitivity, and 80% specificity for the 35 patients used for independent testing. Conclusions This DSS is a promising tool in women diagnosed with OMs at TUS, allowing to predict the individual risk of malignancy, supporting clinical decision making.

Although precision classification is a vital issue for therapy, cancer diagnosis has been shown to have serious constraints. In this paper, we proposed a deep learning model based on gene expression data to perform a pan-cancer classification on 16 cancer types. We used principal component analysis (PCA) to decrease data dimensionality before building a neural network model for pan-cancer prediction. The performance of accuracy was monitored and optimized using the Adam algorithm. We compared the results of the model with a random forest classifier and XGBoost. The results show that the neural network model and random forest achieve high and similar classification performance (neural network mean accuracy: 0.84; random forest mean accuracy: 0.86; XGBoost mean accuracy: 0.90). Thus, we suggest future studies of neural network, random forest and XGBoost models for the detection of cancer in order to identify early treatment approaches to enhance cancer survival.

Abstract Background We aimed to train and test a deep learning classifier to support the diagnosis of coronavirus disease 2019 (COVID-19) using chest x-ray (CXR) on a cohort of subjects from two hospitals in Lombardy, Italy. Methods We used for training and validation an ensemble of ten convolutional neural networks (CNNs) with mainly bedside CXRs of 250 COVID-19 and 250 non-COVID-19 subjects from two hospitals (Centres 1 and 2). We then tested such system on bedside CXRs of an independent group of 110 patients (74 COVID-19, 36 non-COVID-19) from one of the two hospitals. A retrospective reading was performed by two radiologists in the absence of any clinical information, with the aim to differentiate COVID-19 from non-COVID-19 patients. Real-time polymerase chain reaction served as the reference standard. Results At 10-fold cross-validation, our deep learning model classified COVID-19 and non-COVID-19 patients with 0.78 sensitivity (95% confidence interval [CI] 0.74–0.81), 0.82 specificity (95% CI 0.78–0.85), and 0.89 area under the curve (AUC) (95% CI 0.86–0.91). For the independent dataset, deep learning showed 0.80 sensitivity (95% CI 0.72–0.86) (59/74), 0.81 specificity (29/36) (95% CI 0.73–0.87), and 0.81 AUC (95% CI 0.73–0.87). Radiologists’ reading obtained 0.63 sensitivity (95% CI 0.52–0.74) and 0.78 specificity (95% CI 0.61–0.90) in Centre 1 and 0.64 sensitivity (95% CI 0.52–0.74) and 0.86 specificity (95% CI 0.71–0.95) in Centre 2. Conclusions This preliminary experience based on ten CNNs trained on a limited training dataset shows an interesting potential of deep learning for COVID-19 diagnosis. Such tool is in training with new CXRs to further increase its performance.

The exhibition “The Ladies of Art”, held at the Palazzo Reale in Milan in 2021, focused on the history of women artists during the 16th and 17th centuries. As part of the exhibition, a series of diagnostic analyses have been carried out on two paintings, thanks to the collaboration of several research institutions. The considered artworks consist of two paintings on parchments, realized by Giovanna Garzoni at the Savoy Court in the 17th century. Diagnostic analyses were performed using different, complementary, and non-invasive techniques: high-resolution multiband imaging, hyperspectral imaging, optical microscopy, X-ray fluorescence, and FORS spectrometry, combined with advanced post-processing techniques, in order to map and distinguish the pigments and the underdrawing of both the paintings. This research is the first conducted on these paintings and one of the few on the entire works of this important miniaturist. The results showed an incredibly meticulous painting technique, with a detailed metal point underdrawing and painstaking brushstrokes to describe the details with a high degree of realism. Precious materials, such as gold and lapislazuli, were identified and mapped. The findings of this work represent a new contribution of knowledge, which helps to lessen the lack of information for systematic studies on the artistic production of G. Garzoni.

Alzheimer's Disease (AD) is the most common neurodegenerative disease, with 10% prevalence in the elder population. Conventional Machine Learning (ML) was proven effective in supporting the diagnosis of AD, while very few studies investigated the performance of deep learning and transfer learning in this complex task. In this paper, we evaluated the potential of ensemble transfer-learning techniques, pretrained on generic images and then transferred to structural brain MRI, for the early diagnosis and prognosis of AD, with respect to a fusion of conventional-ML approaches based on Support Vector Machine directly applied to structural brain MRI. Specifically, more than 600 subjects were obtained from the ADNI repository, including AD, Mild Cognitive Impaired converting to AD (MCIc), Mild Cognitive Impaired not converting to AD (MCInc), and cognitively-normal (CN) subjects. We used T1-weighted cerebral-MRI studies to train: (1) an ensemble of five transfer-learning architectures pretrained on generic images; (2) a 3D Convolutional Neutral Network (CNN) trained from scratch on MRI volumes; and (3) a fusion of two conventional-ML classifiers derived from different feature extraction/selection techniques coupled to SVM. The AD-vs-CN, MCIc-vs-CN, MCIc-vs-MCInc comparisons were investigated. The ensemble transfer-learning approach was able to effectively discriminate AD from CN with 90.2% AUC, MCIc from CN with 83.2% AUC, and MCIc from MCInc with 70.6% AUC, showing comparable or slightly lower results with the fusion of conventional-ML systems (AD from CN with 93.1% AUC, MCIc from CN with 89.6% AUC, and MCIc from MCInc with AUC in the range of 69.1–73.3%). The deep-learning network trained from scratch obtained lower performance than either the fusion of conventional-ML systems and the ensemble transfer-learning, due to the limited sample of images used for training. These results open new prospective on the use of transfer learning combined with neuroimages for the automatic early diagnosis and prognosis of AD, even if pretrained on generic images.

The “bovine” aortic arch is an anatomic variant consisting in a common origin of the innominate and left carotid artery (CILCA), associated with a greater risk of thoracic aortic diseases (aneurysms and dissections), stroke, and complications after endovascular procedures. CILCA can be detected by visual assessment of computed tomography (CT) chest scans, but it is rarely reported. We developed a deep learning (DL) segmentation-plus-classification system to automatically detect CILCA based on 302 CT studies acquired at 2 centers. One model (3D U-Net) was trained from scratch (supervised by manual segmentation), validated, and tested for the automatic segmentation of the aortic arch and supra-aortic vessels. Three DL architectures (ResNet50, DenseNet-201, and SqueezeNet), pre-trained over millions of common images, were trained, validated, and tested for the automatic classification of CILCA versus non-CILCA, supervised by radiologist’s classification. The 3D U-Net-plus-DenseNet-201 was found to be the best system (Dice index 0.912); its classification performance obtained from internal, independent testing on 126 patients gave a receiver operating characteristic area under the curve of 87.0%, sensitivity 66.7%, specificity 90.5%, positive predictive value 87.5%, negative predictive value 73.1%, positive likelihood ratio 7.0, and negative likelihood ratio 0.4. In conclusion, a combined DL system applied to chest CT scans was developed and proven to be an effective tool to detect individuals with “bovine” aortic arch with a low rate of false-positive findings.

We developed a machine learning model based on radiomics to predict the BI-RADS category of ultrasound-detected suspicious breast lesions and support medical decision-making towards short-interval follow-up versus tissue sampling. From a retrospective 2015–2019 series of ultrasound-guided core needle biopsies performed by four board-certified breast radiologists using six ultrasound systems from three vendors, we collected 821 images of 834 suspicious breast masses from 819 patients, 404 malignant and 430 benign according to histopathology. A balanced image set of biopsy-proven benign (n = 299) and malignant (n = 299) lesions was used for training and cross-validation of ensembles of machine learning algorithms supervised during learning by histopathological diagnosis as a reference standard. Based on a majority vote (over 80% of the votes to have a valid prediction of benign lesion), an ensemble of support vector machines showed an ability to reduce the biopsy rate of benign lesions by 15% to 18%, always keeping a sensitivity over 94%, when externally tested on 236 images from two image sets: (1) 123 lesions (51 malignant and 72 benign) obtained from two ultrasound systems used for training and from a different one, resulting in a positive predictive value (PPV) of 45.9% (95% confidence interval 36.3–55.7%) versus a radiologists’ PPV of 41.5% (p < 0.005), combined with a 98.0% sensitivity (89.6–99.9%); (2) 113 lesions (54 malignant and 59 benign) obtained from two ultrasound systems from vendors different from those used for training, resulting into a 50.5% PPV (40.4–60.6%) versus a radiologists’ PPV of 47.8% (p < 0.005), combined with a 94.4% sensitivity (84.6–98.8%). Errors in BI-RADS 3 category (i.e., assigned by the model as BI-RADS 4) were 0.8% and 2.7% in the Testing set I and II, respectively. The board-certified breast radiologist accepted the BI-RADS classes assigned by the model in 114 masses (92.7%) and modified the BI-RADS classes of 9 breast masses (7.3%). In six of nine cases, the model performed better than the radiologist did, since it assigned a BI-RADS 3 classification to histopathology-confirmed benign masses that were classified as BI-RADS 4 by the radiologist.

Digital breast tomosynthesis (DBT) studies were introduced as a successful help for the detection of calcification, which can be a primary sign of cancer. Expert radiologists are able to detect suspicious calcifications in DBT, but a high number of calcifications with non-malignant diagnosis at biopsy have been reported (false positives, FP). In this study, a radiomic approach was developed and applied on DBT images with the aim to reduce the number of benign calcifications addressed to biopsy and to give the radiologists a helpful decision support system during their diagnostic activity. This allows personalizing patient management on the basis of personalized risk. For this purpose, 49 patients showing microcalcifications on DBT images were retrospectively included, classified by BI-RADS (Breast Imaging-Reporting and Data System) and analyzed. After segmentation of microcalcifications from DBT images, radiomic features were extracted. Features were then selected with respect to their stability within different segmentations and their repeatability in test–retest studies. Stable radiomic features were used to train, validate and test (nested 10-fold cross-validation) a preliminary machine learning radiomic classifier that, combined with BI-RADS classification, allowed a reduction in FP of a factor of 2 and an improvement in positive predictive value of 50%.

We assessed the role of artificial intelligence applied to chest X-rays (CXRs) in supporting the diagnosis of COVID-19. We trained and cross-validated a model with an ensemble of 10 convolutional neural networks with CXRs of 98 COVID-19 patients, 88 community-acquired pneumonia (CAP) patients, and 98 subjects without either COVID-19 or CAP, collected in two Italian hospitals. The system was tested on two independent cohorts, namely, 148 patients (COVID-19, CAP, or negative) collected by one of the two hospitals (independent testing I) and 820 COVID-19 patients collected by a multicenter study (independent testing II). On the training and cross-validation dataset, sensitivity, specificity, and area under the curve (AUC) were 0.91, 0.87, and 0.93 for COVID-19 versus negative subjects, 0.85, 0.82, and 0.94 for COVID-19 versus CAP. On the independent testing I, sensitivity, specificity, and AUC were 0.98, 0.88, and 0.98 for COVID-19 versus negative subjects, 0.97, 0.96, and 0.98 for COVID-19 versus CAP. On the independent testing II, the system correctly diagnosed 652 COVID-19 patients versus negative subjects (0.80 sensitivity) and correctly differentiated 674 COVID-19 versus CAP patients (0.82 sensitivity). This system appears promising for the diagnosis and differential diagnosis of COVID-19, showing its potential as a second opinion tool in conditions of the variable prevalence of different types of infectious pneumonia.

There is no disease-modifying treatment currently available for AD, one of the more impacting neurodegenerative diseases affecting more than 47.5 million people worldwide. The definition of new approaches for the design of proper clinical trials is highly demanded in order to achieve non-confounding results and assess more effective treatment. In this study, a cohort of 200 subjects was obtained from the Alzheimer’s Disease Neuroimaging Initiative. Subjects were followed-up for 24 months, and classified as AD (50), progressive-MCI to AD (50), stable-MCI (50), and cognitively normal (50). Structural T1-weighted MRI brain studies and neuropsychological measures of these subjects were used to train and optimize an artificial-intelligence classifier to distinguish mild-AD patients who need treatment (AD + pMCI) from subjects who do not need treatment (sMCI + CN). The classifier was able to distinguish between the two groups 24 months before AD definite diagnosis using a combination of MRI brain studies and specific neuropsychological measures, with 85% accuracy, 83% sensitivity, and 87% specificity. The combined-approach model outperformed the classification using MRI data alone (72% classification accuracy, 69% sensitivity, and 75% specificity). The patterns of morphological abnormalities localized in the temporal pole and medial-temporal cortex might be considered as biomarkers of clinical progression and evolution. These regions can be already observed 24 months before AD definite diagnosis. The best neuropsychological predictors mainly included measures of functional abilities, memory and learning, working memory, language, visuoconstructional reasoning, and complex attention, with a particular focus on some of the sub-scores of the FAQ and AVLT tests.

The current gold standard for diagnosis of attention deficit/hyperactivity disorder (ADHD) includes subjective measures, such as clinical interview, observation, and rating scales. The significant heterogeneity of ADHD symptoms represents a challenge for this assessment and could prevent an accurate diagnosis. The aim of this work was to investigate the ability of a multi-domain profile of measures, including blood fatty acid (FA) profiles, neuropsychological measures, and functional measures from near-infrared spectroscopy (fNIRS), to correctly recognize school-aged children with ADHD. To answer this question, we elaborated a supervised machine-learning method to accurately discriminate 22 children with ADHD from 22 children with typical development by means of the proposed profile of measures. To assess the performance of our classifier, we adopted a nested 10-fold cross validation, where the original dataset was split into 10 subsets of equal size, which were used repeatedly for training and testing. Each subset was used once for performance validation. Our method reached a maximum diagnostic accuracy of 81% through the combining of the predictive models trained on neuropsychological, FA profiles, and deoxygenated-hemoglobin features. With respect to the analysis of a single-domain dataset per time, the most discriminant neuropsychological features were measures of vigilance, focused and sustained attention, and cognitive flexibility; the most discriminating blood FAs were linoleic acid and the total amount of polyunsaturated fatty acids. Finally, with respect to the fNIRS data, we found a significant advantage of the deoxygenated-hemoglobin over the oxygenated-hemoglobin data in terms of predictive accuracy. These preliminary findings show the feasibility and applicability of our machine-learning method in correctly identifying children with ADHD based on multi-domain data. The present machine-learning classification approach might be helpful for supporting the clinical practice of diagnosing ADHD, even fostering a computer-aided diagnosis perspective.

Subjects with Alzheimer’s disease (AD) show loss of cognitive functions and change in behavioral and functional state affecting the quality of their daily life and that of their families and caregivers. A neuropsychological assessment plays a crucial role in detecting such changes from normal conditions. However, despite the existence of clinical measures that are used to classify and diagnose AD, a large amount of subjectivity continues to exist. Our aim was to assess the potential of machine learning in quantifying this process and optimizing or even reducing the amount of neuropsychological tests used to classify AD patients, also at an early stage of impairment. We investigated the role of twelve state-of-the-art neuropsychological tests in the automatic classification of subjects with none, mild, or severe impairment as measured by the clinical dementia rating (CDR). Data were obtained from the ADNI database. In the groups of measures used as features, we included measures of both cognitive domains and subdomains. Our findings show that some tests are more frequently best predictors for the automatic classification, namely, LM, ADAS-Cog, AVLT, and FAQ, with a major role of the ADAS-Cog measures of delayed and immediate memory and the FAQ measure of financial competency.

Presently, there are no valid biomarkers to identify individuals with eating disorders (ED). The aim of this work was to assess the feasibility of a machine learning method for extracting reliable neuroimaging features allowing individual categorization of patients with ED. Support Vector Machine (SVM) technique, combined with a pattern recognition method, was employed utilizing structural magnetic resonance images. Seventeen females with ED (six with diagnosis of anorexia nervosa and 11 with bulimia nervosa) were compared against 17 body mass index-matched healthy controls (HC). Machine learning allowed individual diagnosis of ED versus HC with an Accuracy ≥ 0.80. Voxel-based pattern recognition analysis demonstrated that voxels influencing the classification Accuracy involved the occipital cortex, the posterior cerebellar lobule, precuneus, sensorimotor/premotor cortices, and the medial prefrontal cortex, all critical regions known to be strongly involved in the pathophysiological mechanisms of ED. Although these findings should be considered preliminary given the small size investigated, SVM analysis highlights the role of well-known brain regions as possible biomarkers to distinguish ED from HC at an individual level, thus encouraging the translational implementation of this new multivariate approach in the clinical practice.