MorphoHELM benchmark reveals classic CV still beats deep learning for cell analysis
No deep learning model outperformed classic computer vision across all settings.
MorphoHELM is a comprehensive open benchmark for evaluating feature extraction methods on Cell Painting microscopy data. It consolidates evaluation standards, extends them to be more robust, and tests a wide range of methods. A key feature is evaluating tasks at different degrees of batch effects to measure how biological signal degrades with noise. The benchmark finds that no existing model outperforms classic computer vision analytic strategies across all settings, which remain the strongest general use-case representations. All datasets, code, and evaluation tools are publicly available.
- MorphoHELM tests 20+ feature extraction methods on Cell Painting data across multiple tasks with controlled batch effects.
- Classic CV (CellProfiler) outperforms deep learning models in general-use representation quality.
- Open-source benchmark with all code, datasets, and evaluation tools available on GitHub.
Why It Matters
Standardized evaluation reveals that fancy deep learning isn't always better—classic CV still wins for drug screening microscopy.