SICAP-MIL: Prostate Whole Slide Images for Histology Classification

SICAP-MIL provides prostate histology whole slide images for weakly supervised classification tasks, as documented in JBHI'21 and CBM'22 publications. Developed by researcher jusiro, the dataset supports medical image analysis and was last updated in October 2024. It is specifically designed for Multiple Instance Learning (MIL) frameworks using slide-level labels.

Use Cases

• Training Multiple Instance Learning (MIL) models using slide-level prostate cancer labels
• Benchmarking weakly supervised classification algorithms on whole slide images
• Developing feature extraction pipelines for histology image analysis

Strengths

• Peer-reviewed validation in JBHI'21 and CBM'22 journals
• MIT license allows for open research and modification
• Specifically formatted for weakly supervised Multiple Instance Learning (MIL) benchmarks

Limitations

• Restricted to prostate histology, limiting generalizability to other tissue types
• Unknown total record count and storage size in provided metadata
• Lack of granular patch-level annotations due to the weakly supervised design

Provenance

Source

jusiro (GitHub), associated with JBHI'21 and CBM'22 publications

Collection Method

Annotated histology slides

Freshness

Last updated October 2024; reflects research state as of the 2021-2022 publications.