Search

Abstract With over 55 million people globally affected by dementia and nearly 10 million new cases reported annually, Alzheimer’s disease is a prevalent and challenging neurodegenerative disorder. Despite significant advancements in machine learning techniques for Alzheimer’s disease detection, the widespread adoption of deep learning models raises concerns about their explainability. The lack of explainability in deep learning models for online handwriting analysis is a critical gap in the literature in the context of Alzheimer’s disease detection. This paper addresses this challenge by interpreting predictions from a Convolutional Neural Network applied to multivariate time series data, generated by online handwriting data associated with continuous loop series handwritten on a graphical tablet. Our explainability methods reveal distinct motor behavior characteristics for healthy individuals and those diagnosed with Alzheimer’s. Healthy subjects exhibited consistent, smooth movements, while Alzheimer’s patients demonstrated erratic patterns marked by abrupt stops and direction changes. This emphasizes the critical role of explainability in translating complex models into clinically relevant insights. Our research contributes to the enhancement of early diagnosis, providing significant and reliable insights to stakeholders involved in patient care and intervention strategies. Our work bridges the gap between machine learning predictions and clinical insights, fostering a more effective and understandable application of advanced models for Alzheimer’s disease assessment.

The water crisis, global warming and climate changes have become recently prominent world issues. Saving and conserving water have become, therefore, an imperative for water resources’ sustainability. In this context, dramatic innovations have come to light, e.g., Precision Agriculture. Numerous research projects on this subject have been conducted, as illustrated not only by the huge number of research works in this topic, but also by the dozens of surveys dedicated to. Surveys regarding smart agriculture are of two categories, surveys on smart agriculture in general in which smart irrigation is a subtopic among others, and surveys dedicated to smart irrigation. Existing surveys, however, suffer from limitations including low paper numbers, linear description without structural classification, and missing key information. To fill this gap, we propose, in this paper, a survey on SI that classifies the large body of literature into categories in a structured way. To this end, we have employed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology to determine the inclusion or exclusion of articles. We have reviewed a total of 610 publications out of which 227 met the inclusion criteria. We have categorized the selected references into three conceptual clusters corresponding to: (1) 41.8% focusing on field measurements, based on IoT sensors, (2) 37% involving Remote Sensing (RS) and (3) 21.1% Artificial Intelligence (AI) methods. Each category is thoroughly described, including the inputs and outputs, along with a description of the various tasks employed in irrigation management. This systematic re-view is expected to serve as a useful reference for research on smart irrigation management methods.

Non-invasive crop analysis through image-based methods holds great promise for applications in plant research, yet accurate and robust trait inference from images remains a critical challenge. Our study investigates the potential of AI model ensembling and hybridization approaches to infer sorghum crop traits from RGB images generated via unmanned aerial vehicle (UAV). In our study, we cultivated 21 sorghum cultivars in two independent seasons (2021 and 2022) with a gradient of fertilizer and water inputs. We collected 470 ground-truth N measurements and captured corresponding RGB images with a drone-mounted camera. We computed five RGB vegetation indices, employed several ML models such as MLR, MLP, and various CNN architectures (season 2021), and compared their prediction accuracy for N-inference on the independent test set (season 2022). We assessed strategies that leveraged both deep and handcrafted features, namely hybridized and ensembled AI architectures. Our approach considered two different datasets collected during the two seasons (2021 and 2022), with the training set from the first season only. This allowed for testing of the models’ robustness, particularly their sensitivity to concept drifts, in the independent season (2022), which is fundamental for practical agriculture applications. Our findings underscore the superiority of hybrid and ensembled AI algorithms in these experiments. The MLP + CNN-VGG16 combination achieved the best accuracy (R2 = 0.733, MAE = 0.264 N% on an independent dataset). This study emphasized that carefully crafted AI-based models applied to RGB images can achieve robust trait prediction with accuracies comparable to the similar phenotyping tasks using more complex (multi- and hyper-spectral) sensors presented in the current literature.

Building upon the recent advances and successes in the application of deep learning to the medical field, we propose in this work a new approach to detect and classify early-stage Alzheimer patients using online handwriting (HW) loop patterns. To cope with the lack of training data prevalent in the tasks of classification of neuro-degenerative diseases from behavioral data, we investigate several data augmentation techniques. In this respect, compared to the traditional data augmentation techniques proposed for HW-based Parkinson detection, we investigate a variant of Generative Adversarial Networks (GANs), DoppelGANger, especially tailored for times series and hence suitable for synthesizing realistic online handwriting sequences. Based on a 1D-Convolutional Neural Network (1D-CNN) to perform Alzheimer classification, we show, on a real dataset related to HW and Alzheimer, that our DoppelGANger-based augmentation model allow the CNN to significantly outperform both the current state of the art and the other data augmentation techniques.

Deep neural networks have been widely used in several complex tasks such as robotics, self-driving cars, medicine, etc. However, they have recently shown to be vulnerable in uncertain environments where inputs are noisy. As a consequence, the robustness of neural networks has become an essential property for their application in critical systems. Robustness is the capacity to take the same decision even when inputs are disturbed under different types of perturbations, including adversarial attacks. The great difficulty today is providing a formal guarantee of robustness, which is the context of this paper. To do so, abstract interpretation, a popular state-of-the-art method, consisting of converting the layers of the neural network into abstract layers, has been recently proposed. An abstract layer can act on a geometric abstract object or shape comprising implicitly an infinite number of inputs rather than an individual input. In this paper, we propose a new mathematical formulation of an abstract transformer to convert a LeakyReLU activation layer to an abstract layer. Moreover, we implement and integrate our transformer into the ERAN tool. For validation, we assess the performance of our transformer according to the LeakyReLU hyperparameter, and we study the robustness of the neural network according to the input perturbation intensity. Our approach is evaluated on three different datasets: MNIST, Fashion and a robotic dataset. The obtained results demonstrate the efficacy of our abstract transformer in terms of mathematical formulation and implementation.

Artificial Neural Networks (ANNs) have amassed unprecedented success in information processing ranging from image recognition to time series prediction. The success can largely be attributed to the availability of large datasets for training and the increased complexity of the models. Unfortunately, for some applications only a limited amount of samples is available for training. Fewer training samples increases the risk of over-fitting and poor generalization especially in high complexity models. Moreover, complex models with a large number of trainable parameters require more energy to train and optimize compared to simpler ones. In this paper, to the best of our knowledge, we propose the first use of ANNs for Early Stage Alzheimer Disease classification (ES-AD) from the handwriting (HW). We propose using a framework for building Recurrent Neural Networks (RNNs) known as Reservoir Computing (RC), both numerically and experimentally, that simplifies training by optimizing the output layer only. We also propose the Bidirectional Long Term Short Term (BiLSTM) and Convolutional Neural Network (CNN) methods for comparison. For a fairer comparison, we not only consider the accuracies but also the energy costs incurred to obtain the respective accuracies in order to assess the accuracy-efficiency trade-off. Our numerical and experimental results show that RC yields a classification accuracy of 85%, which is 3% worse than that of BiLSTM and 2% better than that of CNN, at a relatively lower training and significantly lower inference costs. We hope that our findings highlight the importance of examining the accuracy-efficiency trade-off of various models in the community in order to reduce the overall impact of ANNs training on the environment.

Most of recent methods for action/activity recognition, usually based on static classifiers, have achieved improvements by integrating context of local interest point (IP) features such as spatiotemporal IPs by characterising their neighbourhood under different scales. In this study, the authors propose a new approach that explicitly models the sequential aspect of activities. First, a sliding window segmentation technique splits the video stream into overlapping short segments. Each window is characterised by a local bag of words of IPs encoded by motion information. A first‐layer support vector machine provides for each window a vector of conditional class probabilities that summarises all discriminant information that is relevant for sequence recognition. The sequence of these stochastic vectors is then fed to a hidden conditional random field for inference at the sequence level. They also show how their approach can be naturally extended to the problem of conjoint segmentation and recognition of a sequence of action classes within a continuous video stream. They have tested their model on various human action and activity datasets and the obtained results compare favourably with current state of the art.

We present an autonomous assistive robotic system for human activity recognition from video sequences. Due to the large variability inherent to video capture from a non-fixed robot (as opposed to a fixed camera), as well as the robot's limited computing resources, implementation has been guided by robustness to this variability and by memory and computing speed efficiency. To accommodate motion speed variability across users, we encode motion using dense interest point trajectories. Our recognition model harnesses the dense interest point bag-of-words representation through an intersection kernel-based SVM that better accommodates the large intra-class variability stemming from a robot operating in different locations and conditions. To contextually assess the engine as implemented in the robot, we compare it with the most recent approaches of human action recognition performed on public datasets (non-robot-based), including a novel approach of our own that is based on a two-layer SVM-hidden conditional random field sequential recognition model. The latter's performance is among the best within the recent state of the art. We show that our robot-based recognition engine, while less accurate than the sequential model, nonetheless shows good performances, especially given the adverse test conditions of the robot, relative to those of a fixed camera.

Photoplethysmography (PPG) is a non-invasive and cost-efficient optical technique used to assess blood volume variations in the microcirculation. PPG technology is widely used in a variety of wearable sensors to investigate the cardiovascular system. Recent studies have demonstrated the utility of PPG analysis for carrying out large-scale screening to prevent and detect diabetes. However, most of these studies require feature extraction and/or several pre-processing steps. Over the past few years, the advent of deep learning has significantly impacted the analysis of biomedical signals. Despite their success in other fields, however, very few studies have focused on the application of deep learning to raw PPG signals for detecting diabetes. Existing studies have proposed large models trained on large amounts of data. In this paper, we present a Light CNN-based model for screening the presence of type 2 diabetes using a single raw pulse extracted from photoplethysmographic signals. In addition to the baseline architecture, we evaluate different model architectures that take as input age and biological sex or PPG handcrafted features. Furthermore, we apply transfer learning to all the tested architectures to evaluate the effectiveness of harnessing pre-trained models in detecting diabetes. We tested a model pre-trained on a general PPG shape dataset and another model pre-trained on a dataset containing hypertension PPG signals. Our model scored an AUC of 75.5 when trained with raw PPG waves, age, and biological sex without applying transfer learning, which is competitive with current state of the art.

Photopletysmography (PPG) is a non-invasive and well known technology that enables the recording of the digital volume pulse (DVP). Although PPG is largely employed in research, several aspects remain unknown. One of these is represented by the lack of information about how many waveform classes best express the variability in shape. In the literature, it is common to classify DVPs into four classes based on the dicrotic notch position. However, when working with real data, labelling waveforms with one of these four classes is no longer straightforward and may be challenging. The correct identification of the DVP shape could enhance the precision and the reliability of the extracted bio markers. In this work we proposed unsupervised machine learning and deep learning approaches to overcome the data labelling limitations. Concretely we performed a K-medoids based clustering that takes as input 1) DVP handcrafted features, 2) similarity matrix computed with the Derivative Dynamic Time Warping and 3) DVP features extracted from a CNN AutoEncoder. All the cited methods have been tested first by imposing four medoids representative of the Dawber classes, and after by automatically searching four clusters. We then searched the optimal number of clusters for each method using silhouette score, the prediction strength and inertia. To validate the proposed approaches we analyse the dissimilarities in the clinical data related to obtained clusters.

(1) Background: Diabetes mellitus (DM) is a chronic, metabolic disease characterized by elevated levels of blood glucose. Recently, some studies approached the diabetes care domain through the analysis of the modifications of cardiovascular system parameters. In fact, cardiovascular diseases are the first leading cause of death in diabetic subjects. Thanks to their cost effectiveness and their ease of use, electrocardiographic (ECG) and photoplethysmographic (PPG) signals have recently been used in diabetes detection, blood glucose estimation and diabetes-related complication detection. This review’s aim is to provide a detailed overview of all the published methods, from the traditional (non machine learning) to the deep learning approaches, to detect and manage diabetes using PPG and ECG signals. This review will allow researchers to compare and understand the differences, in terms of results, amount of data and complexity that each type of approach provides and requires. (2) Method: We performed a systematic review based on articles that focus on the use of ECG and PPG signals in diabetes care. The search was focused on keywords related to the topic, such as “Diabetes”, “ECG”, “PPG”, “Machine Learning”, etc. This was performed using databases, such as PubMed, Google Scholar, Semantic Scholar and IEEE Xplore. This review’s aim is to provide a detailed overview of all the published methods, from the traditional (non machine learning) to the deep learning approaches, to detect and manage diabetes using PPG and ECG signals. This review will allow researchers to compare and understand the differences, in terms of results, amount of data and complexity that each type of approach provides and requires. (3) Results: A total of 78 studies were included. The majority of the selected studies focused on blood glucose estimation (41) and diabetes detection (31). Only 7 studies focused on diabetes complications detection. We present these studies by approach: traditional, machine learning and deep learning approaches. (4) Conclusions: ECG and PPG analysis in diabetes care showed to be very promising. Clinical validation and data processing standardization need to be improved in order to employ these techniques in a clinical environment.

Hand-vein biometrics as a high-security pattern has received more and more attention. One of the open issues in hand-vein verification is the lack of robustness against image quality degradation, which may comprise the verification accuracy. To achieve robust verification, vein feature extraction approaches, especially vein texture segmentation, have been extensively investigated. In recent years, deep neural networks have achieved promising results in medical image segmentation and have been brought into vein verification, but current solutions suffer from two challenges for vein segmentation: 1) lacking the labeling data, which is expensive to obtain and 2) the incorrect label data obtained by manual labeling scheme or automatic labeling scheme may strongly influence parameters when the network is trained, which may degrade the verification performance. This paper proposes an iterative deep belief network (DBN) to extract vein features based on the initial label data, which are automatically generated using a very limited a priori knowledge and iteratively corrected by our DBN. First, a known handcrafted vein image segmentation technique is employed to automatically label vein pixel and background pixel. A training dataset is constructed based on the patches centered on the labeled pixels. Second, a DBN is trained on the resulting database to predict the probability of each pixel to belong to be a vein pixel given a patch centered on it. The vein patterns are segmented using a probability threshold of 0.5. The resulting vein features are employed to reconstruct the training dataset, based on which the network is retrained. During the iterative procedure, the incorrect labels of training data are statistically corrected, which enables DBN to effectively learn what a finger-vein pattern is by learning the difference between vein patterns and background ones. The experimental results on two public hand-vein databases show a significant improvement in terms of hand-vein verification accuracy.