[ICCV 2025] AbdomenAtlas 3.0 (9,262 CT volumes + medical reports). These “superhuman” reports are more accurate, detailed, standardized, and generated faster than traditional human-made reports.

[IMAVIS] Official implementation of "ASF-YOLO: A Novel YOLO Model with Attentional Scale Sequence Fusion for Cell Instance Segmentation".

Liver Lesion Segmentation with 2D Unets

Official website of our paper: Applications of Deep Learning in Fundus Images: A Review. Newly-released datasets and recently-published papers will be updated regularly.

Patho-GAN: interpretation + medical data augmentation. Code for paper work "Explainable Diabetic Retinopathy Detection and Retinal Image Generation"

Segmentation of skin cancers on ISIC 2017 challenge dataset.

TextSCF: LLM-Enhanced Image Registration Model

[BIBM 2025] The official code for "Only Positive Cases: 5-fold High-order Attention Interaction Model for Skin Segmentation Derived Classification".

Official code for "DermSynth3D: Synthesis of in-the-wild Annotated Dermatology Images". A data generation pipeline for creating photorealistic in-the-wild synthetic dermatalogical data with rich multi-task annotations for various skin-analysis tasks.

PyTorch Implementation of Small Lesion Segmentation in Brain MRIs with Subpixel Embedding (ORAL, MICCAIW 2021)

DL tool for white matter hyperintensities segmentation

Official PyTorch Implementation of ModDrop++ [MICCAI 2022 (early accept)]. A simple yet effective approach to tackle missing-modality problem for multi-modality medical imaging data.

DeepWMH: Annotation-free white matter hyperintensity (WMH) lesion segmentation tool using deep learning.

Sequence Adaptive Multimodal SEGmentation (SAMSEG)

Official development code of the Automatic Scoring of Atopic Dermatitis (ASCORAD) by Legit.Health 🩺🤖

A comprehensive platform for analyzing pulmonary parenchyma lesions on chest CT.

[BSPC] The official code for "MHorUNet: High-order spatial interaction UNet for skin lesion segmentation".

Fully automatic skin lesion segmentation using the Berkeley wavelet transform and UNet algorithm.

A production-grade deep learning system for automated skin lesion classification using the HAM10000 dataset. This system provides training, evaluation, and real-time inference capabilities for detecting seven types of skin lesions.

The official code for "LiteMamba-Bound: A lightweight Mamba-based model with boundary-aware and normalized active contour loss for skin lesion segmentation"