CAMEL: A Weakly Supervised Learning Framework for Histopathology Image Segmentation

TL;DR

CAMEL introduces a weakly supervised framework that leverages image-level labels and multiple instance learning to effectively segment histopathology images, reducing the need for labor-intensive pixel-level annotations.

Contribution

The paper presents a novel weakly supervised learning method that automatically generates pixel-level labels from image-level labels for histopathology image segmentation.

Findings

Achieves comparable results to fully supervised methods on CAMELYON16 and colorectal datasets.

Reduces annotation effort by using only image-level labels.

Demonstrates the generality of the label enrichment approach for histopathology analysis.

Abstract

Histopathology image analysis plays a critical role in cancer diagnosis and treatment. To automatically segment the cancerous regions, fully supervised segmentation algorithms require labor-intensive and time-consuming labeling at the pixel level. In this research, we propose CAMEL, a weakly supervised learning framework for histopathology image segmentation using only image-level labels. Using multiple instance learning (MIL)-based label enrichment, CAMEL splits the image into latticed instances and automatically generates instance-level labels. After label enrichment, the instance-level labels are further assigned to the corresponding pixels, producing the approximate pixel-level labels and making fully supervised training of segmentation models possible. CAMEL achieves comparable performance with the fully supervised approaches in both instance-level classification and pixel-level segmentation on CAMELYON16 and a colorectal adenoma dataset. Moreover, the generality of the automatic labeling methodology may benefit future weakly supervised learning studies for histopathology image analysis.