A5036911172- AI in cancer detection
- Human Pose and Action Recognition
- Multimodal Machine Learning Applications
- Genomic variations and chromosomal abnormalities
- Genomics and Chromatin Dynamics
- Radiomics and Machine Learning in Medical Imaging
- Advanced Image and Video Retrieval Techniques
- COVID-19 diagnosis using AI
- Advanced Image Processing Techniques
- Cell Image Analysis Techniques
- Biomedical Text Mining and Ontologies
- Generative Adversarial Networks and Image Synthesis
- Video Surveillance and Tracking Methods
- Video Analysis and Summarization
- Medical Image Segmentation Techniques
- Hedgehog Signaling Pathway Studies
Zhejiang University of Science and Technology
2024-2025
Westlake University
2022-2024
Zhejiang University
2024
Accurate image classification and retrieval are of importance for clinical diagnosis treatment decision-making. The recent contrastive language-image pre-training (CLIP) model has shown remarkable proficiency in understanding natural images. Drawing inspiration from CLIP, pathology-dedicated CLIP (PathCLIP) been developed, utilizing over 200,000 text pairs training. While the performance PathCLIP is impressive, its robustness under a wide range corruptions remains unknown. Therefore, we...
Synthetic data generation emerges as a strategy to mitigate scarcity in digital pathology, where complicated tissue and cellular features are correlated with cancer diagnosis. The synthesis of such visuals, however, suffers from limited inter class diversity annotations. Current methodologies struggle capturing the broad spectrum pathology features, causing unpredictable objects defected fidelity. Moreover, discrepancies image resolution across developmental operational phases can amplify...
Whole slide images are the foundation of digital pathology for diagnosis and treatment carcinomas. Writing reports is laborious error-prone inexperienced pathologists. To reduce workload improve clinical automation, we investigate how to generate given whole images. On data end, curated largest WSI-text dataset (TCGA-PathoText). In specific, collected nearly 10000 high-quality pairs visual-language models by recognizing cleaning which narrate diagnostic slides in TCGA. model propose multiple...
Accurate image classification and retrieval are of importance for clinical diagnosis treatment decision-making. The recent contrastive language-image pretraining (CLIP) model has shown remarkable proficiency in understanding natural images. Drawing inspiration from CLIP, PathCLIP is specifically designed pathology analysis, utilizing over 200,000 text pairs training. While the performance impressive, its robustness under a wide range corruptions remains unknown. Therefore, we conduct an...
The increasing variety and quantity of tagged multimedia content on platforms such as TikTok provides an opportunity to advance computer vision modeling. We have curated a distinctive dataset 283,582 unique video clips categorized under 386 hashtags relating modern human actions. release this valuable resource for building domain-specific foundation models movement modeling tasks action recognition. To validate dataset, which we name TikTokActions, perform two sets experiments. First,...
Whole slide imaging is routinely adopted for carcinoma diagnosis and prognosis. Abundant experience required pathologists to achieve accurate reliable diagnostic results of whole images (WSI). The huge size heterogeneous features WSIs make the workflow pathological reading extremely time-consuming. In this paper, we propose a novel framework (WSI-VQA) interpret by generative visual question answering. WSI-VQA shows universality reframing various kinds slide-level tasks in question-answering...
Karyotyping is an important procedure to assess the possible existence of chromosomal abnormalities. However, because non-rigid nature, chromosomes are usually heavily curved in microscopic images and such deformed shapes hinder chromosome analysis for cytogeneticists. In this paper, we present a self-attention guided framework erase curvature chromosomes. The proposed extracts spatial information local textures preserve banding patterns regression module. With complementary from bent...