Photoplethysmography (PPG) is a non-invasive technique for detecting heart rate (HR) but is often hampered by noise, which affects its reliability in cardiac monitoring applications like heart rate variability (HRV) and blood pressure measurement. To overcome this, a Wearable LOw-Cost PPG acquisition deVicE (WeLOVE) was developed, and variational mode decomposition combined with principal component analysis (VMD-PCA) is proposed to estimate HR and other cardiovascular parameters from PPG signals. The accuracy of WeLOVE was validated against the Equivital wireless physiological monitoring system (EQO2) under various breathing conditions. The mean absolute error (MAE) of HR between the PPG signal and ECG signal was 7.8±3.25 bpm during normal breathing and 8.86±4.45 bpm during slow breathing across subjects. The root mean square error (RMSE) was 13.46±5.42 bpm for normal breathing and 17.16±8.34 bpm for slow breathing across subjects. The average HR measured by ECG was 71.77±8.15 bpm for normal breathing and 68.85±9.5 bpm for slow breathing, while PPG readings were 65.61±6.34 bpm and 65.14±7.64 bpm, respectively, across subjects. These results demonstrate that the WeLOVE device, combined with the VMD-PCA method, offers significantly accurate heart rate (HR) measurements while effectively mitigating motion artifacts. The robustness of this approach ensures reliable performance even in dynamic conditions, making it particularly well-suited for real-time applications in cardiac monitoring. This capability is essential for continuous health tracking, wearable medical devices, and remote patient monitoring, where precise and artifact-resistant HR estimation is crucial for timely and effective clinical decision-making.

ABSTRACT Obstructive sleep apnea (OSA) is a common and serious sleep disorder characterized by periodic interruptions in breathing lasting more than 10 s (apnea episodes) during sleep. OSA significantly affects quality of life and overall health, highlighting the critical need for an accurate and timely diagnosis. Polysomnography (PSG) is the standard diagnostic technique for OSA, involving the collection of respiratory, oxygen saturation, biometric, and physiological signals. However, manual analysis of these extensive sleep recordings by medical professionals is labor‐intensive and time‐consuming. To address this challenge, we propose a Siamese Network‐based Self‐Supervised Learning (SSSL) model for the automatic identification of SA episodes from single‐channel electrocardiogram (ECG) signals. Unlike conventional self‐supervised methods, our approach does not require a momentum encoder, large batch sizes, or negative‐positive pair construction. The model is evaluated using the PhysioNet Apnea‐ECG database and employs a two‐stage training strategy. In the first stage, the encoder is trained on unlabeled data to learn robust signal representations. In the second stage, the pre‐trained encoder and classifier are fine‐tuned using labelled data for optimal classification performance. The proposed model achieved high accuracy of , , and when fine‐tuned with , , and of the labelled training data, respectively, for the classification per segment. These results demonstrate the model's effectiveness in both offline and online diagnostic settings, outperforming state‐of‐the‐art methods.

Early identification and management of Diabetic Foot Ulcers (DFU) are critical for preventing severe complications, particularly in the Indian subcontinent where DFU prevalence is high. This paper introduces UlcerMTL, a multi-task learning (MTL) model designed to simultaneously perform binary classification (ulcer vs. healthy skin) and semantic segmentation of the ulcer boundaries. Our approach leverages a shared SegFormer-B2 encoder to extract common features, which are then passed to task-specific classification and segmentation heads. This enables effective specialization for each task while benefiting from shared early representations. We evaluated our model on a newly collected dataset comprising 1,534 diabetic foot ulcer patches and 1,533 healthy skin patches from patients in the Indian subcontinent. Our model achieves a classification accuracy of 99.7%, with a segmentation performance yielding an Intersection over Union (IoU) of 0.709 and a Dice score of 0.829. Compared to single-task baselines, our approach offers significant improvements, demonstrating strong potential for clinical deployment in DFU screening and management.

Obstructive sleep apnea (OSA) is an extremely severe condition. At present, the conventional polysomnography (PSG) test is utilized to treat OSA using several physiological signals, such as electroencephalogram (EEG), electrocardio-gram (ECG), and the oxygen level in their blood. During the PSG test, the patient is required to wear many sensors while sleeping. This method is unquestionably complex, expensive, and can cause discomfort to the patients. In addition, it is more appropriate to use single-lead ECG signals for wearable mobile devices due to their compatibility with noninvasive requirements and hardware limitations. In this research, a deep learning model (DLM) using smoothed wavelet spectrograms (SWS) of ECG signals is proposed for the automatic classi-fication of OSA. The SWSs are smoothened using Savitzky-Golay (S-G) filter. Then, these SWSs are provided as input to the designed DLM called WAvelet SPectrogram-based Convo-lutional Neural Network (WASPCN-Net) and pre-trained Res-Net50 model. The WASPCN-Net model obtained an accuracy of 87.25%, sensitivity of 78.97%, and specificity of 92.35% with SWS using a 10-fold cross-validation approach, which is superior to many state-of-the-art techniques. Further, we also obtained the performance on the pre-trained ResNet-50 model and received an accuracy of 86.92%, sensitivity of 81.74%, and specificity of 90.12%. The proposed WASPCN-Net is more accurate, simple, fast, and robust than the Res-Net50 model because it requires relatively few tunable learning parameters.

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that is usually diagnosed between the ages of one and three. It is characterized by developmental issues and repetitive behaviours. One of the important social skills, i.e., Joint Attention (JA), involves developing a shared focus of attention with another person. Children with ASD often lack JA skills, which can impede their ability to develop social communication skills later in life. This makes early intervention critical. Previous research has explored various techniques for teaching JA skill training, but few have utilized immersive Augmented Reality (AR) and Virtual Reality (VR) based devices for JA skill training. Additionally, there is limited work in literature exploring the physiological effect of JA training via using immersive AR and VR devices. This paper addresses these gaps by introducing a novel JA training platform that utilizes immersive AR and VR devices for JA skill training. This platform enables participants to interact with it using their eye gaze. To validate the acceptance of the developed platforms, we conducted experiments on ASD (5) and Neurotypical (NT) (10) participants. To quantify the participant's task performance while interacting with these platforms, we have used behavioural (time duration to register a response) and physiological parameters (Beats per minute (BPM)). The ASD group took a longer time for response registration than the NT on both AR and VR platforms (mean duration in sec, for ASD (AR/VR): 34.5/12.5; for NT (AR/VR): 8.8/4.22). Also, the physiological parameter BPM showed a similar trend, which was higher in ASD in comparison to NT for both platforms (BPM, for ASD (AR/VR): 100.49/90.27; for NT (AR/VR): 87.39/86.60). The increase in cardiac activity, as quantified by BPM values for ASD, gives us an impression of the sensory sensitivities in the autistic group that lead to physiological arousal and thereby interfere with their focusing capability, resulting in delayed response. This study emphasizes the importance of monitoring physiological responses of participants during JA training. It also highlights the difficulties faced by ASD participants during these trainings in immersive AR and VR environments.

This article introduces a novel shallow 3-D self-supervised tensor neural network in quantum formalism for volumetric segmentation of medical images with merits of obviating training and supervision. The proposed network is referred to as the 3-D quantum-inspired self-supervised tensor neural network (3-D-QNet). The underlying architecture of 3-D-QNet is composed of a trinity of volumetric layers, viz., input, intermediate, and output layers interconnected using an S -connected third-order neighborhood-based topology for voxelwise processing of 3-D medical image data, suitable for semantic segmentation. Each of the volumetric layers contains quantum neurons designated by qubits or quantum bits. The incorporation of tensor decomposition in quantum formalism leads to faster convergence of network operations to preclude the inherent slow convergence problems faced by the classical supervised and self-supervised networks. The segmented volumes are obtained once the network converges. The suggested 3-D-QNet is tailored and tested on the BRATS 2019 Brain MR image dataset and the Liver Tumor Segmentation Challenge (LiTS17) dataset extensively in our experiments. The 3-D-QNet has achieved promising dice similarity (DS) as compared with the time-intensive supervised convolutional neural network (CNN)-based models, such as 3-D-UNet, voxelwise residual network (VoxResNet), Dense-Res-Inception Net (DRINet), and 3-D-ESPNet, thereby showing a potential advantage of our self-supervised shallow network on facilitating semantic segmentation.

Coronavirus disease has caused unprecedented chaos across the globe causing potentially fatal pneumonia, since the beginning of 2020. Researchers from different communities are working in conjunction with front-line doctors and policymakers to better understand the disease. The key to prevent the spread is a rapid diagnosis, prioritized isolation, and fastidious contact tracing. Recent studies have confirmed the presence of underlying patterns on chest CT for patients with COVID-19. We present a completely automated framework to detect COVID-19 using chest CT scans, only needing a small number of training samples. We present a few-shot learning technique based on the Triplet network in comparison to the conventional deep learning techniques which require a substantial amount of training examples. We used 140 chest CT images for training and the rest for testing from a total of 2482 images for both COVID-19 and non-COVID-19 cases from a publicly available dataset. The model trained with chest CT images achieves an AUC of 0.94, separates the two classes into distinct clusters; thereby giving correct prediction accuracy on the evaluation dataset.

Stress due to water deficiency in plants can significantly lower the agricultural yield. It can affect many visible plant traits such as size and surface area, the number of leaves and their color, etc. In recent years, computer vision-based plant phenomics has emerged as a promising tool for plant research and management. Such techniques have the advantage of being non-destructive, non-evasive, fast, and offer high levels of automation. Pulses like chickpeas play an important role in ensuring food security in poor countries owing to their high protein and nutrition content. In the present work, we have built a dataset comprising of two varieties of chickpea plant shoot images under different moisture stress conditions. Specifically, we propose a BAT optimized ResNet-18 model for classifying stress induced by water deficiency using chickpea shoot images. BAT algorithm identifies the optimal value of the mini-batch size to be used for training rather than employing the traditional manual approach of trial and error. Experimentation on two crop varieties (JG and Pusa) reveals that BAT optimized approach achieves an accuracy of 96% and 91% for JG and Pusa varieties that is better than the traditional method by 4%. The experimental results are also compared with state of the art CNN models like Alexnet, GoogleNet, and ResNet-50. The comparison results demonstrate that the proposed BAT optimized ResNet-18 model achieves higher performance than the comparison counterparts.

The deep learning models, which include attention mechanisms, are shown to enhance the performance and efficiency of the various computer vision tasks such as pattern recognition, object detection, face recognition, etc. Although the visual attention mechanism is the source of inspiration for these models, recent attention models consider 'attention' as a pure machine vision optimization problem, and visual attention remains the most neglected aspect. Therefore, this paper presents a collaborative human and machine attention module which considers both visual and network's attention. The proposed module is inspired by the dorsal (`where') pathways of visual processing and can be integrated with any convolutional neural network (CNN) model. First, the module computes the spatial attention map from the input feature maps, which is then combined with the visual attention maps. The visual attention maps are created using eye-fixations obtained by performing an eye-tracking experiment with human participants. The visual attention map covers the highly salient and discriminating image regions as humans tend to focus on such regions, whereas the other relevant image regions are processed by spatial attention map. The combination of these two maps results in the finer refinement in feature maps, resulting in improved performance. The comparative analysis reveals that our model not only shows significant improvement over the baseline model but also outperforms the other models. We hope that our findings using a collaborative human-machine attention module will be helpful in other computer vision tasks as well.

Hazy environment attenuates the scene radiance and causes difficulty in distinguishing the color and texture of the scene. A crucial step in dehazing is the recovery of the global air-light vector. Traditional methods usually interpret the RGB value of the brightest region in haze images as the air-light. In this letter, a new prior called `color constancy prior' has been proposed to improve the robustness of air-light estimation when varicolored illumination exists. The prior utilizes the statistical observation that distant scenery objects become the most haze-opaque due to the pixel escalation towards the higher intensity side. The comparative evaluation on a variety of haze images manifests that the proposed prior perform better than existing air-light recovery methods and can be used for subsequent dehazing applications.

Images captured in underwater (UW) are often disturbed with several kind of degradation such as low visibility, non-uniform color cast, haze, and blurriness. To date, most UW image restoration methods have ignored the effects of sensor blur and noise. Therefore, in this paper, we propose a novel three stage algorithm for visibility recovery in UW images by considering both sensor blur and noise. In the first stage, blind deconvolution is used for the estimation of an unknown point spread function (PSF). In the second stage, a new prior called weighted median channel prior (WMCP) is used for the estimation of scene depth and background light. In the third stage, a color balancing (CB) module is adopted to minimize the effect of non-uniform color cast. Experimental results manifest that the proposed algorithm is effective and has the character of visibility improvement, and color correction than previous state-of-the-art methods.