A5070095990- Radiomics and Machine Learning in Medical Imaging
- Advanced X-ray and CT Imaging
- Medical Imaging and Analysis
- COVID-19 diagnosis using AI
- Medical Image Segmentation Techniques
- AI in cancer detection
- Lung Cancer Diagnosis and Treatment
- Artificial Intelligence in Healthcare and Education
- Scoliosis diagnosis and treatment
- Cutaneous Melanoma Detection and Management
- Brain Tumor Detection and Classification
- Advanced Neural Network Applications
- Medical Imaging Techniques and Applications
- Dental Radiography and Imaging
- Digital Imaging for Blood Diseases
- Cardiac Imaging and Diagnostics
- Cardiac Valve Diseases and Treatments
- Radiation Dose and Imaging
- Data-Driven Disease Surveillance
- 3D Shape Modeling and Analysis
- Spinal Fractures and Fixation Techniques
- Anomaly Detection Techniques and Applications
- ECG Monitoring and Analysis
- Chronic Obstructive Pulmonary Disease (COPD) Research
- Generative Adversarial Networks and Image Synthesis
Advanced Imaging Research (United States)
2023-2025
Duke University
2023-2025
Pratt Institute
2024
Institute for Social and Environmental Research-Nepal
2019-2022
University of Girona
2019-2020
Since the COVID-19 pandemic, several research studies have proposed Deep Learning (DL)-based automated detection, reporting high cross-validation accuracy when classifying patients from normal or other common Pneumonia. Although reported outcomes are very in most cases, these results were obtained without an independent test set a separate data source(s). DL models likely to overfit training distribution sets not utilized prone learn dataset-specific artifacts rather than actual disease...
In medical imaging, harmonization plays a crucial role in reducing variability arising from diverse imaging devices and protocols. Patient images obtained under different computed tomography (CT) scan conditions may show varying performance when analyzed using an artificial intelligence model or quantitative assessment. This necessitates the need for harmonization. Virtual trial (VIT) through digital simulation can be used to develop assess effectiveness of models minimize data variability....
ABSTRACT A large number of studies in the past months have proposed deep learning-based Artificial Intelligence (AI) tools for automated detection COVID-19 using publicly available datasets Chest X-rays (CXRs) or CT scans training and evaluation. Most these report high accuracy when classifying patients from normal other commonly occurring pneumonia cases. However, results are often obtained on cross-validation without an independent test set coming a separate dataset biases such as two...
Automatic segmentation of brain Magnetic Resonance Imaging (MRI) images is one the vital steps for quantitative analysis further inspection. In this paper, NeuroNet has been adopted to segment tissues (white matter (WM), grey (GM) and cerebrospinal fluid (CSF)) which uses Residual Network (ResNet) in encoder Fully Convolution (FCN) decoder. To achieve best performance, various hyper-parameters have tuned, while, network parameters (kernel bias) were initialized using pre-trained model....
Virtual Imaging Trials, known as VITs, provide a computational substitute for clinical trials. These traditional trials tend to be sluggish, costly, and frequently deficient in definitive evidence, all the while subjecting participants ionizing radiation. Our VIT platform meticulously mimics essential components of imaging process, encompassing everything from virtual patients scanners simulated readers. Within scope this intended research, we aim authenticate our trial by duplicating...
We develop the XCAT series of phantoms for medical imaging research. The model different individuals over various ages, heights, and weights, but a current drawback is they do not include small intestine variability. Each phantom has derived from common anatomical template due to difficulty fitting regular tubular patient segmentations. Building upon previous work, we software pipeline add realistic variability in by generating surface models with random length diameter fit within...
BACKGROUND: Lung cancer's high mortality rate can be mitigated by early detection, which is increasingly reliant on artificial intelligence (AI) for diagnostic imaging. However, the performance of AI models contingent upon datasets used their training and validation. METHODS: This study developed validated DLCSD-mD LUNA16-mD utilizing Duke Cancer Screening Dataset (DLCSD), encompassing over 2,000 CT scans with more than 3,000 annotations. These were rigorously evaluated against internal...
Virtual Imaging Trials (VIT) offer a cost-effective and scalable approach for evaluating medical imaging technologies. Computational phantoms, which mimic real patient anatomy physiology, play central role in VIT. However, the current libraries of computational phantoms face limitations, particularly terms sample size diversity. Insufficient representation population hampers accurate assessment technologies across different groups. Traditionally, were created by manual segmentation, is...
2D echocardiography is the most common imaging modality for cardiovascular diseases. The portability and relatively low-cost nature of Ultrasound (US) enable US devices needed performing to be made widely available. However, acquiring interpreting cardiac images operator dependent, limiting its use only places where experts are present. Recently, Deep Learning (DL) has been used in automated view classification, structure function assessment. Although these recent works show promise...
Correct evaluation and treatment of Scoliosis require accurate estimation spinal curvature. Current gold standard is to manually estimate Cobb Angles in X-ray images which time consuming has high inter-rater variability. We propose an automatic method with a novel framework that first detects vertebrae as objects followed by landmark detector estimates the 4 corners each vertebra separately. are calculated using slope obtained from predicted landmarks. For inference on test data, we perform...
Deep learning methods have performed superiorly to segment organs of interest from Computed Tomography images than traditional methods. However, the trained models do not generalize well at inference phase, and manual validation correction are feasible for large-scale studies. Therefore, automatic detect segmentation failure crucial Computer Aided Diagnosis systems. In this work, we present an quality control method that can be used reject poor segmentation. We register new test cases...
The credibility of AI models in medical imaging is often challenged by reproducibility issues and obscured clinical insights, a reality highlighted during the COVID-19 pandemic many reports near-perfect artificial intelligence (AI) that all failed to generalize. To address these concerns, we propose virtual trial framework, employing diverse collection images are both simulated. In this study, serves as case example unveil intrinsic extrinsic factors influencing performance. Our findings...
Accurate 3D modeling of human organs plays a crucial role in building computational phantoms for virtual imaging trials. However, generating anatomically plausible reconstructions organ surfaces from computed tomography scans remains challenging many structures the body. This challenge is particularly evident when dealing with large intestine. In this study, we leverage recent advancements geometric deep learning and denoising diffusion probabilistic models to refine segmentation results We...
Automatic segmentation of skin lesion is considered a crucial step in Computer Aided Diagnosis (CAD) for melanoma diagnosis. Despite its significance, remains challenging task due to their diverse color, texture, and indistinguishable boundaries forms an open problem. Through this study, we present new automatic semantic network robust named Dermoscopic Skin Network (DSNet). In order reduce the number parameters make lightweight, used depth-wise separable convolution lieu standard project...
Automatic segmentation of brain Magnetic Resonance Imaging (MRI) images is one the vital steps for quantitative analysis further inspection. In this paper, NeuroNet has been adopted to segment tissues (white matter (WM), grey (GM) and cerebrospinal fluid (CSF)) which uses Residual Network (ResNet) in encoder Fully Convolution (FCN) decoder. To achieve best performance, various hyper-parameters have tuned, while, network parameters (kernel bias) were initialized using pre-trained model....
Many published studies use deep learning models to predict COVID-19 from chest x-ray (CXR) images, often reporting high performances. However, the do not generalize well on independent external testing. Common limitations include lack of medical imaging data and disease labels, leading training small datasets or drawing classes different institutions. To address these concerns, we designed an validation study classifiers for in CXR images including XCAT phantoms as well. We hypothesize that...