Multiple Instance Captioning: Learning Representations from Histopathology Textbooks and Articles
This provides a foundational dataset for computational pathology, enabling improved representation learning across multiple sub-tasks in the field.
The authors introduced ARCH, a computational pathology dataset with dense diagnostic and morphological descriptions for various stains, tissue types, and pathologies, showing it rivals MS-COCO Captions in intrinsic dimensionality and that pre-training on it yields representations that transfer better to pathology tasks than ImageNet or other methods.
We present ARCH, a computational pathology (CP) multiple instance captioning dataset to facilitate dense supervision of CP tasks. Existing CP datasets focus on narrow tasks; ARCH on the other hand contains dense diagnostic and morphological descriptions for a range of stains, tissue types and pathologies. Using intrinsic dimensionality estimation, we show that ARCH is the only CP dataset to (ARCH-)rival its computer vision analog MS-COCO Captions. We conjecture that an encoder pre-trained on dense image captions learns transferable representations for most CP tasks. We support the conjecture with evidence that ARCH representation transfers to a variety of pathology sub-tasks better than ImageNet features or representations obtained via self-supervised or multi-task learning on pathology images alone. We release our best model and invite other researchers to test it on their CP tasks.