Breast cancer remains a major concern for women’s lives worldwide and serves as evidence of the need for better classification strategies according to severity. Computer-aided diagnosis (CADx) powered by explainable artificial intelligence (XAI) offers a promising solution by minimizing diagnostic errors and fostering trust through a more transparent decision-making process. As XAI evolves, it plays a crucial role in increasing the interpretability of AI-driven diagnostics, particularly in distributed healthcare systems. XAI explains the model’s prediction, making clinicians more confident in accepting clinical outcomes. Accordingly, this study provides a comparative analysis of multiple deep learning models for breast cancer identification based on a publicly available dataset of 780 ultrasound images with their masks. Explanation of the classification result is then provided using the Grad-CAM method which improves the interpretability and tractability of the models. The proposed method lets the models explain their decisions visually, using heatmaps that show what part of an image contributes to the predictions in a valuable way when studying medical images. Obtained results demonstrate the XAI’s transformative potential in medical imaging, paving the way for more reliable, scalable, and efficient diagnostic tools. Also, providing a critical comparison of various types of deep learning models for breast cancer identification, the study underlines the advantages and limitations of the different architectures in solving the task.

Monitoring physiological parameters such as respiratory rate (f$_{R}$) is essential for diagnosing and managing various pathological conditions. Thermal imaging offers a promising contactless alternative to traditional methods, which often rely on partially invasive sensors or obtrusive wearable systems. However, existing approaches for f$_{R}$ estimation from thermal signals typically require extensive pre-processing and manual or semi-automatic region-of-interest (ROI) tracking, limiting their practical applicability. This study proposes a deep learning-based method for estimating f$_{R}$ directly from thermal videos, eliminating the need for complex pre-processing and ROI tracking. A 3D Convolutional Neural Network (3D-CNN) is developed to operate on raw thermal video data. To address challenges related to small datasets, the model is trained using data augmentation and transfer learning from synthetic datasets. Experimental results demonstrate that the proposed approach achieves a validation $R^{2}$ score of approximately 0.61 on both pre-processed and raw thermal videos. By simplifying the workflow, this method holds promise for enhancing the feasibility of thermal imaging in real-world applications, such as remote healthcare and driver monitoring in automotive applications.

Stroke, if not lethal, is a primary cause of disability. Early assessment of markers of recovery can allow personalized interventions; however, it is difficult to deliver indexes in the acute phase able to predict recovery. In this perspective, evaluation of electrical brain activity may provide useful information. A machine learning approach was explored here to predict post-stroke recovery relying on multi-channel electroencephalographic (EEG) recordings of few minutes performed at rest. A data-driven model, based on partial least square (PLS) regression, was trained on 19-channel EEG recordings performed within 10 days after mono-hemispheric stroke in 101 patients. The band-wise (delta: 1–4[Formula: see text]Hz, theta: 4–7[Formula: see text]Hz, alpha: 8–14[Formula: see text]Hz and beta: 15–30[Formula: see text]Hz) EEG effective powers were used as features to predict the recovery at 6 months (based on clinical status evaluated through the NIH Stroke Scale, NIHSS) in an optimized and cross-validated framework. In order to exploit the multimodal contribution to prognosis, the EEG-based prediction of recovery was combined with NIHSS scores in the acute phase and both were fed to a nonlinear support vector regressor (SVR). The prediction performance of EEG was at least as good as that of the acute clinical status scores. A posteriori evaluation of the features exploited by the analysis highlighted a lower delta and higher alpha activity in patients showing a positive outcome, independently of the affected hemisphere. The multimodal approach showed better prediction capabilities compared to the acute NIHSS scores alone ([Formula: see text] versus [Formula: see text], AUC = 0.80 versus AUC = 0.70, [Formula: see text]). The multimodal and multivariate model can be used in acute phase to infer recovery relying on standard EEG recordings of few minutes performed at rest together with clinical assessment, to be exploited for early and personalized therapies. The easiness of performing EEG may allow such an approach to become a standard-of-care and, thanks to the increasing number of labeled samples, further improving the model predictive power.

Children with insufficient exposure to language during critical developmental periods in infancy are at risk for cognitive, language, and social deficits [55]. This is especially difficult for deaf infants, as more than 90% are born to hearing parents with little sign language experience [48]. We created an integrated multi-agent system involving a robot and virtual human designed to augment language exposure for 6-12 month old infants. Human-machine design for infants is challenging, as most screen-based media are unlikely to support learning in [33]. While presently, robots are incapable of the dexterity and expressiveness required for signing, even if it existed, developmental questions remain about the capacity for language from artificial agents to engage infants. Here we engineered the robot and avatar to provide visual language to effect socially contingent human conversational exchange. We demonstrate the successful engagement of our technology through case studies of deaf and hearing infants.

We present dialogue management routines for a system to engage in multiparty agent-infant interaction. The ultimate purpose of this research is to help infants learn a visual sign language by engaging them in naturalistic and socially contingent conversations during an early-life critical period for language development (ages 6 to 12 months) as initiated by an artificial agent. As a first step, we focus on creating and maintaining agent-infant engagement that elicits appropriate and socially contingent responses from the baby. Our system includes two agents, a physical robot and an animated virtual human. The system's multimodal perception includes an eye-tracker (measures attention) and a thermal infrared imaging camera (measures patterns of emotional arousal). A dialogue policy is presented that selects individual actions and planned multiparty sequences based on perceptual inputs about the baby's internal changing states of emotional engagement. The present version of the system was evaluated in interaction with 8 babies. All babies demonstrated spontaneous and sustained engagement with the agents for several minutes, with patterns of conversationally relevant and socially contingent behaviors. We further performed a detailed case-study analysis with annotation of all agent and baby behaviors. Results show that the baby's behaviors were generally relevant to agent conversations and contained direct evidence for socially contingent responses by the baby to specific linguistic samples produced by the avatar. This work demonstrates the potential for language learning from agents in very young babies and has especially broad implications regarding the use of artificial agents with babies who have minimal language exposure in early life.

Interpersonal communication and relationships rely on the continuous assessment of the psychophysiological state of the interlocutor. The thermal infrared imaging has been demonstrated to be a reliable tool for the non-invasive contactless evaluation of vital signs, psychophysiological responses and emotional states. The use of this technique is quickly spreading in many fields, from social and developmental psychology to the psychometrics; or from the continuous monitoring of vital signs and stress up to the human-machine interaction. The state of the art of thermal infrared imaging in psychophysiology and in the assessment of emotional states is discussed to provide insights about its potentialities and limits, with special emphasis to the still open issues for image processing and real-time data analysis.

Nailfold capillaroscopy (NC) is a non-invasive imaging technique employed to assess the condition of blood capillaries in the nailfold, and is known to be effective particularly for early detection of scleroderma spectrum disorders and evaluation of Raynaud's phenomenon. Manual inspection of NC images can be aided by a computerised system avoiding the inherent ambiguity present in human judgment and improving the diagnosis speed. For such an automated analysis, image enhancement is typically the first step. The performance of an employed image enhancement algorithm is crucial due to its influence on subsequent algorithms. In this paper, we aim to provide a comparative evaluation of different image enhancement techniques for nailfold capillaroscopy (NC) images. In particular, we evaluate the performance of ten image enhancement/noise removal techniques for NC images as a pre-cursor to edge detection aimed at identifying capillaries. Results on a variety of NC images show that bilateral filters and enhancers, non local means and anisotropic diffusion provide the best image quality for this task.

We have developed a novel method to measure human cardiac pulse at a distance. It is based on the information contained in the thermal signal emitted from major superficial vessels. This signal is acquired through a highly sensitive thermal imaging system. Temperature on the vessel is modulated by pulsative blood flow. To compute the frequency of modulation (pulse), we extract a line-based region along the vessel. Then, we apply Fast Fourier Transform (FFT) to individual points along this line of interest to capitalize on the pulse propagation effect. Finally, we use an adaptive estimation function on the average FFT outcome to quantify the pulse. We have tested the accuracy of our method on 5 subjects with highly successful results. The technology is expected to find applications among others in sustained physiological monitoring of cardiopulmonary diseases, sport training, sleep studies, and psychophysiology (polygraph).