A5086797294- Radiomics and Machine Learning in Medical Imaging
- COVID-19 diagnosis using AI
- Artificial Intelligence in Healthcare and Education
- Explainable Artificial Intelligence (XAI)
- AI in cancer detection
- Machine Learning in Healthcare
- Retinal Imaging and Analysis
- Lung Cancer Diagnosis and Treatment
- Glioma Diagnosis and Treatment
- Advanced Neural Network Applications
- Robotics and Automated Systems
- Pneumonia and Respiratory Infections
- Digital and Cyber Forensics
- Retinal Diseases and Treatments
- Digital Radiography and Breast Imaging
- Domain Adaptation and Few-Shot Learning
- Advanced Malware Detection Techniques
- Generative Adversarial Networks and Image Synthesis
- EEG and Brain-Computer Interfaces
- Hand Gesture Recognition Systems
- Phonocardiography and Auscultation Techniques
- Advanced Memory and Neural Computing
- Network Security and Intrusion Detection
- Gaze Tracking and Assistive Technology
- Medical Imaging and Analysis
Vels University
2025
Carnegie Mellon University
2025
Massachusetts General Hospital
2020-2023
Athinoula A. Martinos Center for Biomedical Imaging
2020-2023
National Institute of Biomedical Imaging and Bioengineering
2022
National Institutes of Health
2022
Weatherford College
2022
Harvard University
2020-2021
Shiv Nadar University
2020-2021
Harvard University Press
2020
Purpose To evaluate the trustworthiness of saliency maps for abnormality localization in medical imaging. Materials and Methods Using two large publicly available radiology datasets (Society Imaging Informatics Medicine–American College Radiology Pneumothorax Segmentation dataset Radiological Society North America Pneumonia Detection Challenge dataset), performance eight commonly used map techniques were quantified regard to (a) utility (segmentation detection), (b) sensitivity model weight...
Purpose To develop an automated measure of COVID-19 pulmonary disease severity on chest radiographs for longitudinal tracking and outcome prediction. Materials Methods A convolutional Siamese neural network–based algorithm was trained to output a (pulmonary x-ray [PXS] score), using weakly supervised pretraining approximately 160 000 anterior-posterior images from CheXpert transfer learning 314 frontal patients with COVID-19. The evaluated internal external test sets different hospitals (154...
Deep learning (DL) models have provided state-of-the-art performance in various medical imaging benchmarking challenges, including the Brain Tumor Segmentation (BraTS) challenges. However, task of focal pathology multi-compartment segmentation (e.g., tumor and lesion sub-regions) is particularly challenging, potential errors hinder translating DL into clinical workflows. Quantifying reliability model predictions form uncertainties could enable review most uncertain regions, thereby building...
Saliency maps have become a widely used method to make deep learning models more interpretable by providing post-hoc explanations of classifiers through identification the most pertinent areas input medical image. They are increasingly being in imaging provide clinically plausible for decisions neural network makes. However, utility and robustness these visualization has not yet been rigorously examined context imaging. We posit that trustworthiness this requires 1) localization utility, 2)...
ABSTRACT Purpose To develop an automated measure of COVID-19 pulmonary disease severity on chest radiographs (CXRs), for longitudinal evaluation and clinical risk stratification. Materials Methods A convolutional Siamese neural network-based algorithm was trained to output a anterior-posterior CXRs (pulmonary x-ray (PXS) score), using weakly-supervised pretraining ~160,000 images from CheXpert transfer learning 314 patients with COVID-19. The evaluated internal external test sets different...
To tune and test the generalizability of a deep learning-based model for assessment COVID-19 lung disease severity on chest radiographs (CXRs) from different patient populations. A published convolutional Siamese neural network-based previously trained hospitalized patients with was tuned using 250 outpatient CXRs. This produces quantitative measure (pulmonary x-ray (PXS) score). The evaluated CXRs 4 sets, including 3 United States (patients at an academic medical center (N = 154), community...
Saliency maps have become a widely used method to assess which areas of the input image are most pertinent prediction trained neural network. However, in context medical imaging, there is no study our knowledge that has examined efficacy these techniques and quantified them using overlap with ground truth bounding boxes. In this work, we explored credibility various existing saliency map methods on RSNA Pneumonia dataset. We found GradCAM was sensitive model parameter label randomization,...
ABSTRACT Purpose To improve and test the generalizability of a deep learning-based model for assessment COVID-19 lung disease severity on chest radiographs (CXRs) from different patient populations. Materials Methods A published convolutional Siamese neural network-based previously trained hospitalized patients with was tuned using 250 outpatient CXRs. This produces quantitative measure (pulmonary x-ray (PXS) score). The evaluated CXRs four sets, including 3 United States (patients at an...
Abstract Model brittleness is a key concern when deploying deep learning models in real-world medical settings. A model that has high performance at one dataset may suffer significant decline tested on different datasets. While pooling datasets from multiple hospitals and re-training provide straightforward solution, it often infeasible compromise patient privacy. An alternative approach to fine-tune the subsequent after training original dataset. Notably,this degrades datasets, phenomenon...
Abstract The global COVID-19 pandemic has disrupted patient care delivery in healthcare systems world-wide. For providers to better allocate their resources and improve the for patients with severe disease, it is valuable be able identify those who are at higher risk clinical complications. This may help optimize workflow more efficiently scarce medical resources. To this end, imaging shows great potential artificial intelligence (AI) algorithms have been developed assist diagnosing...
Model brittleness is a key concern when deploying deep learning models in real-world medical settings. A model that has high performance at one institution may suffer significant decline tested other institutions. While pooling datasets from multiple institutions and retraining provide straightforward solution, it often infeasible compromise patient privacy. An alternative approach to fine-tune the on subsequent after training original institution. Notably, this degrades institution,...
Model brittleness is a primary concern when deploying deep learning models in medical settings owing to inter-institution variations, like patient demographics and intra-institution variation, such as multiple scanner types. While simply training on the combined datasets fraught with data privacy limitations, fine-tuning model subsequent institutions after it original institution results decrease performance dataset, phenomenon called catastrophic forgetting. In this paper, we investigate...
Abstract Heritability of common eye diseases and ocular traits are relatively high. Here, we develop an automated algorithm to detect genetic relatedness from color fundus photographs (FPs). We estimated the degree shared ancestry amongst individuals in UK Biobank using KING software. A convolutional Siamese neural network-based was trained output a measure 7224 pairs (3612 related 3612 unrelated) FPs. The model achieved high performance for prediction relatedness; when computed Euclidean...
BrainGate is a cutting edge brain machine interface (BMI) system that enables individuals with severe motor disabilities to control external devices using their neural activity. This paper provides comprehensive overview of the technology, its underlying principles, technical implementation, applications, and current state research. The also discusses challenges future directions for BMI technologies in general.
While success of Deep Learning (DL) in automated diagnosis can be transformative to the medicinal practice especially for people with little or no access doctors, its widespread acceptability is severely limited by inherent black-box decision making and unsafe failure modes. saliency methods attempt tackle this problem non-medical contexts, their apriori explanations do not transfer well medical usecases. With study we validate a model design element agnostic both architecture complexity...
Deep learning (DL) models have provided state-of-the-art performance in various medical imaging benchmarking challenges, including the Brain Tumor Segmentation (BraTS) challenges. However, task of focal pathology multi-compartment segmentation (e.g., tumor and lesion sub-regions) is particularly challenging, potential errors hinder translating DL into clinical workflows. Quantifying reliability model predictions form uncertainties could enable review most uncertain regions, thereby building...