search

mahmoodlab / HEST

HEST: Bringing Spatial Transcriptomics and Histopathology together - NeurIPS 2024
Other
164 stars 12 forks source link

Why don't we use all the samples to construct the benchmark? #27

Closed huangtinglin closed 3 months ago

huangtinglin commented 3 months ago

Thanks for the valuable work and dataset, which inspires me a lot.

I found that only 86 samples were used for benchmarking, while the whole dataset included around 1K samples. I wonder if we can construct a larger benchmark by incorporating the whole HEST-1K dataset.

guillaumejaume commented 3 months ago

Hi, thanks for your interest in our work. It is definitively possible to construct a larger benchmark. HEST-Benchmark was built based on human cancer samples, whereas HEST-1k includes mouse samples, healthy cases, and samples with other diseases. Still, we encourage other researchers to explore other tasks that could be developed from HEST-1k, such as enlarging HEST-Benchmark.

huangtinglin commented 3 months ago

That makes sense! Thank you for answering.