Proportion Estimation by Masked Learning from Label Proportion

TL;DR

This paper introduces a novel masked learning approach for estimating PD-L1 tumor cell positivity rates from pathological images, requiring minimal cell-level annotations and effectively handling data imbalance.

Contribution

It presents a new proportion estimation method combining tumor cell detection, masked learning from label proportions, and a weighted focal loss to improve accuracy with limited annotations.

Findings

Achieved superior performance on clinical data

Effectively handles data imbalance

Requires minimal cell-level annotation

Abstract

The PD-L1 rate, the number of PD-L1 positive tumor cells over the total number of all tumor cells, is an important metric for immunotherapy. This metric is recorded as diagnostic information with pathological images. In this paper, we propose a proportion estimation method with a small amount of cell-level annotation and proportion annotation, which can be easily collected. Since the PD-L1 rate is calculated from only `tumor cells' and not using `non-tumor cells', we first detect tumor cells with a detection model. Then, we estimate the PD-L1 proportion by introducing a masking technique to `learning from label proportion.' In addition, we propose a weighted focal proportion loss to address data imbalance problems. Experiments using clinical data demonstrate the effectiveness of our method. Our method achieved the best performance in comparisons.