A5069518008- Human Pose and Action Recognition
- COVID-19 diagnosis using AI
- Advanced Neural Network Applications
- AI in cancer detection
- ECG Monitoring and Analysis
- Gait Recognition and Analysis
- Heart Rate Variability and Autonomic Control
- Anomaly Detection Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Radiomics and Machine Learning in Medical Imaging
- Advanced Vision and Imaging
- Emotion and Mood Recognition
- Face and Expression Recognition
- Medical Image Segmentation Techniques
- Domain Adaptation and Few-Shot Learning
- Non-Invasive Vital Sign Monitoring
- Machine Learning in Healthcare
- Robotics and Sensor-Based Localization
- Phonocardiography and Auscultation Techniques
- EEG and Brain-Computer Interfaces
- Explainable Artificial Intelligence (XAI)
- Augmented Reality Applications
- Computer Graphics and Visualization Techniques
- Brain Tumor Detection and Classification
- Hand Gesture Recognition Systems
Toronto Metropolitan University
2016-2025
Metropolitan University
2024
University of Windsor
2009-2023
Southeast University
2023
Shenzhen University
2023
Fudan University
2023
The University of Tokyo
2023
We propose a generalized focal loss function based on the Tversky index to address issue of data imbalance in medical image segmentation. Compared commonly used Dice loss, our achieves better trade off between precision and recall when training small structures such as lesions. To evaluate function, we improve attention U-Net model by incorporating an pyramid preserve contextual features. experiment BUS 2017 dataset ISIC 2018 where lesions occupy 4.84% 21.4% images area segmentation accuracy...
Detection of Alzheimer's Disease (AD) from neuroimaging data such as MRI through machine learning have been a subject intense research in recent years. Recent success deep computer vision progressed further. However, common limitations with algorithms are reliance on large number training images, and requirement careful optimization the architecture networks. In this paper, we attempt solving these issues transfer learning, where state-of-the-art architectures VGG Inception initialized...
Local Interpretable Model-Agnostic Explanations (LIME) is a popular technique used to increase the interpretability and explainability of black box Machine Learning (ML) algorithms. LIME typically creates an explanation for single prediction by any ML model learning simpler interpretable (e.g., linear classifier) around through generating simulated data instance random perturbation, obtaining feature importance applying some form selection. While similar local algorithms have gained...
Detection of Alzheimer's disease (AD) from neuroimaging data such as MRI through machine learning has been a subject intense research in recent years. The success deep computer vision progressed research. However, common limitations with algorithms are reliance on large number training images, and the requirement careful optimization architecture networks. In this paper, we attempt solving these issues transfer learning, where state-of-the-art VGG is initialized pre-trained weights benchmark...
Electrocardiogram (ECG) is an authoritative source to diagnose and counter critical cardiovascular syndromes such as arrhythmia myocardial infarction (MI). Current machine learning techniques either depend on manually extracted features or large complex deep networks which merely utilize the 1D ECG signal directly. Since intelligent multimodal fusion can perform at state-of-the-art level with efficient network, therefore, in this paper, we propose two computationally frameworks for heart...
<p>We propose a generalized focal loss function based on the Tversky index to address issue of data imbalance in medical image segmentation. Compared commonly used Dice loss, our functionachievesa better trade offbetween precision and recall when training small structures such as lesions. To evaluate function, we improve attention U-Net model by incorporating an pyramid preserve contextual features. We experiment BUS 2017 dataset ISIC 2018 where lesions occupy 4.84% 21.4% images area...
<p>Local Interpretable Model-Agnostic Explanations (LIME) is a popular technique used to increase the interpretability and explainability of black box Machine Learning (ML) algorithms. LIME typically generates an explanation for single prediction by any ML model learning simpler interpretable (e.g. linear classifier) around through generating simulated data instance random perturbation, obtaining feature importance applying some form selection. While similar local algorithms have...
Abstract Virtual reality (VR) for mental health promotion remains understudied in low-income humanitarian settings. We examined the effectiveness of VR reducing depression with urban refugee youth Kampala, Uganda. This randomized controlled trial assessed alone (Arm 1), followed by Group Problem Management Plus (GPM+) 2) and a control group 3), peer-driven convenience sample aged 16–25 Kampala. The primary outcome, depression, was measured Patient Health Questionnaire-9. Secondary outcomes...
Multimodal fusion frameworks for Human Action Recognition (HAR) using depth and inertial sensor data have been proposed over the years. In most of existing works, is performed at a single level (feature or decision level), missing opportunity to fuse rich mid-level features necessary better classification. To address this shortcoming, in paper, we propose three novel deep multilevel multimodal (M2) capitalize on different strategies various stages leverage superiority fusion. At input,...
Recently, Graph Convolutional Network(GCN) methods for skeleton-based action recognition have achieved great success due to their ability preserve structural information of the skeleton. However, these abandon in classification stage by employing traditional fully-connected layers and softmax classifier, leading sub-optimal performance. In this work, a novel Networks-Hidden conditional Random Field (GCN-HCRF) model is proposed solve problem. The method combines GCN with HCRF retain human...
Physiological signals are the most reliable form of for emotion recognition, as they cannot be controlled deliberately by subject. Existing review papers on recognition based physiological surveyed only regular steps involved in workflow such pre-processing, feature extraction, and classification. While these important steps, required any signal processing application. Emotion poses its own set challenges that very to address a robust system. Thus, bridge gap existing literature, this paper,...
Local Interpretable Model-Agnostic Explanations (LIME) is a popular technique used to increase the interpretability and explainability of black box Machine Learning (ML) algorithms. LIME typically generates an explanation for single prediction by any ML model learning simpler interpretable (e.g. linear classifier) around through generating simulated data instance random perturbation, obtaining feature importance applying some form selection. While similar local algorithms have gained...
Convolutional Neural Network (CNN) provides leverage to extract and fuse features from all layers of its architecture.However, extracting fusing intermediate different CNN structure is still uninvestigated for Human Action Recognition (HAR) using depth inertial sensors.To get maximum benefit accessing the CNN's layers, in this paper, we propose novel Multistage Gated Average Fusion (MGAF) network which extracts fuses our computationally efficient (GAF) network, a decisive integral element...
Convolutional Neural Networks (CNNs) are successful deep learning models in the field of computer vision. To get maximum advantage CNN model for Human Action Recognition (HAR) using inertial sensor data, this paper, we use four types spatial domain methods transforming data to activity images, which then utilized a novel fusion framework. These images Signal Images (SI), Gramian Angular Field (GAF) Images, Markov Transition (MTF) and Recurrence Plot (RP) Images. Furthermore, creating...
<p>Detection of Alzheimer's Disease (AD) from neuroimaging data such as MRI through machine learning have been a subject intense research in recent years. Recent success deep computer vision progressed further. However, common limitations with algorithms are reliance on large number training images, and requirement careful optimization the architecture networks. In this paper, we attempt solving these issues transfer learning, where state-of-the-art architectures VGG Inception...
Stress can affect a person's performance and health positively negatively. A lot of the relaxation methods have been suggested to reduce amount stress. This study used virtual reality (VR) video games alleviate Physiological signals captured from Electrocardiogram (ECG), galvanic skin response (GSR), respiration (RESP) were determine if subject was stressed or relaxed. Time frequency domain features then extracted evaluate stress levels. Frequency such as low-frequency (LF), high-frequency...
Traditionally, convolutional neural networks need large amounts of data labelled by humans to train. Self supervision has been proposed as a method dealing with small data. The aim this study is determine whether self can increase classification performance on COVID-19 CT scan dataset. This also aims the strategy, targeted supervision, viable option for imaging A total 10 experiments are run comparing different strategy perform significantly better than their non-self supervised...
<p>We propose a generalized focal loss function based on the Tversky index to address issue of data imbalance in medical image segmentation. Compared commonly used Dice loss, our functionachievesa better trade offbetween precision and recall when training small structures such as lesions. To evaluate function, we improve attention U-Net model by incorporating an pyramid preserve contextual features. We experiment BUS 2017 dataset ISIC 2018 where lesions occupy 4.84% 21.4% images area...
The labor market is undergoing a rapid artificial intelligence (AI) revolution. There currently limited empirical scholarship that focuses on how AI adoption affects employment opportunities and work environments in ways shape worker health, safety, well-being equity. In this article, we present an agenda to guide research examining the implications of intersection between health. To build agenda, full day meeting was organized attended by 50 participants including researchers from diverse...
Electrocardiogram (ECG) is an attractive option to assess stress in serious virtual reality (VR) applications due its noninvasive nature. However, the existing machine learning (ML) models perform poorly. Moreover, studies only a binary assessment, while develop more engaging biofeedback-based application, multilevel assessment necessary. Existing annotate and classify single experience (e.g., watching VR video) level, which again prevents design of dynamic experiences where real-time...