A5102909439- COVID-19 diagnosis using AI
- AI in cancer detection
- Radiomics and Machine Learning in Medical Imaging
- Artificial Intelligence in Healthcare and Education
- Video Surveillance and Tracking Methods
- Medical Image Segmentation Techniques
- Advanced Neural Network Applications
- Colorectal Cancer Screening and Detection
- Image and Object Detection Techniques
- Face recognition and analysis
- Advanced Vision and Imaging
- Industrial Vision Systems and Defect Detection
- Image Retrieval and Classification Techniques
- Advanced X-ray and CT Imaging
- Colorectal Cancer Surgical Treatments
- Machine Learning in Healthcare
- Gastrointestinal disorders and treatments
- Visual Attention and Saliency Detection
- Gastrointestinal Bleeding Diagnosis and Treatment
- Anomaly Detection Techniques and Applications
- Hepatocellular Carcinoma Treatment and Prognosis
- Brain Tumor Detection and Classification
- Retinal Imaging and Analysis
- Gastrointestinal Tumor Research and Treatment
- Advanced Image and Video Retrieval Techniques
VinUniversity
2021-2023
Vinh University
2022
Korea Advanced Institute of Science and Technology
2013
Abstract Saliency methods, which produce heat maps that highlight the areas of medical image influence model prediction, are often presented to clinicians as an aid in diagnostic decision-making. However, rigorous investigation accuracy and reliability these strategies is necessary before they integrated into clinical setting. In this work, we quantitatively evaluate seven saliency including Grad-CAM, across multiple neural network architectures using two evaluation metrics. We establish...
Anomaly detection is an important application in large-scale industrial manufacturing. Recent methods for this task have demonstrated excellent accuracy but come with a latency trade-off. Memory based approaches dominant performances like PatchCore or Coupled-hypersphere-based Feature Adaptation (CFA) require external memory bank, which significantly lengthens the execution time. Another approach that employs Reversed Distillation (RD) can perform well while maintaining low latency. In...
Current research on deep learning for medical image segmentation exposes their limitations in either global semantic information or local contextual information. To tackle these issues, a novel network named SegTransVAE is proposed this paper. built upon encoder-decoder architecture, exploiting transformer with the variational autoencoder (VAE) branch to reconstruct input images jointly segmentation. best of our knowledge, first method combining success CNN, transformer, and VAE. Evaluation...
Early detection of liver malignancy based on medical image analysis plays a crucial role in patient prognosis and personalized treatment. This task, however, is challenging due to several factors, including data scarcity limited training samples. paper presents study three important aspects radiomics feature from multiphase computed tomography (CT) for classifying hepatocellular carcinoma (HCC) other focal lesions: wavelet-transformed extraction, relevant selection, features-based...
Abstract Saliency methods, which “explain” deep neural networks by producing heat maps that highlight the areas of medical image influence model prediction, are often presented to clinicians as an aid in diagnostic decision-making. Although many saliency methods have been proposed for imaging interpretation, rigorous investigation accuracy and reliability these strategies is necessary before they integrated into clinical setting. In this work, we quantitatively evaluate seven...
Automatic extraction of medical conditions from free-text radiology reports is critical for supervising computer vision models to interpret images. In this work, we show that radiologists labeling significantly disagree with corresponding chest X-ray images, which reduces the quality report labels as proxies image labels. We develop and evaluate methods produce have better agreement Our best performing method, called VisualCheXbert, uses a biomedically-pretrained BERT model directly map...
Clinical decision support tools can improve diagnostic performance or reduce variability, but they are also subject to post-deployment underperformance. Although using AI in an assistive setting offsets many concerns with autonomous medicine, systems that present all predictions equivalently fail protect against key safety concerns. We design a pipeline supports the model ecosystem of models, integrating disagreement prediction, clinical significance categorization, and prediction quality...
Object re-identification (ReID) is prone to errors under variations in scale, illumination, complex background, and object occlusion scenarios. To overcome these challenges, attention mechanisms are employed focus on the object's characteristics, thereby extracting better discriminative features. This paper introduces a local-global vision transformer (LoGoViT) for by learning hierarchical-level representation from fine-grained (local) general (global) context It comprises two components:...
Due to the shortage of experienced endoscopists, Computer-Aided Diagnosis (CAD) systems for colonoscopy have recently attracted many research interests. There exist several public polyp segmentation datasets, giving way adoptions domain adaptation methods address shift in distributions. Current frameworks often comprise (i) a discriminator trained with an adversarial loss and (ii) image-translation network. complexity networks, such are generally hard train achieve satisfactory results....
Deep metric learning (or simply called learning) uses the deep neural network to learn representation of images, leading widely used in many applications, e.g. image retrieval and face recognition. In approaches, proxy anchor takes advantage proxy-based pair-based approaches enable fast convergence time robustness noisy labels. However, training anchor, selecting hyperparameter margin is important achieve a good performance. This selection requires expertise time-consuming. paper proposes...
Image segmentation plays a crucial role in many clinical applications, including disease diagnosis and monitoring. Current state-of-the-art approaches use deep neural networks that are trained on their target tasks by minimizing loss function. Class imbalance is one of the major challenges these face, where object significantly underrepresented. Compound functions incorporate binary cross-entropy (BCE) Dice among most prominent to address this issue. However, determining contribution each...
Accurate alignment of multi-session medical imaging is essential to the analysis disease progression. By comparing magnetic resonance (MRI) data captured before and after a course neoadjuvant chemoradiation (nCRT) treatment, physicians are able evaluate tumor response for further treatment disease. However, rectal MRI in often misaligned not guaranteed have one-to-one correspondence, making it challenging observe tumor. To address this issue, we propose an unsupervised learning based volume...
Shape recognition of solder ball bumps in a BGA (Ball Grid Array) is an important issue flip chip bonding technology. In particular, the semiconductor industry has required faster and more accurate inspection micron-size as density balls increased dramatically. The difficulty this comes from specular reflection on metal ball. very realproblem for computer vision systems. Specular appears, disappears, or changes its image abruptly due to tiny movementson behalf viewer. This paper presents...
As an improvement over the standard simple online real-time tracking (SORT) method, DeepSORT introduces a cascade matching mechanism to track objects during certain period of occlusion, effectively reducing number identity (ID) switches. However, lacks capability lost-identities recovery, which enables robustness and performance in face recognition systems. To address issue, we propose novel multi-face named ReSORT, that can recover lost identities. Our method removes block extends...
Many clinical AI-based decision support tools have been shown to increase diagnostic performance or reduce variability, but as with any human-AI collaboration system, these potential benefits also come risks of post-deployment underperformance. Though the use AI in an assistive setting offsets many and ethical concerns autonomous medicine, systems that present all positive negative model predictions equivalently, regardless accuracy, urgency, likelihood disagreement, fail protect against...