Conditional Training with Bounding Map for Universal Lesion Detection

TL;DR

This paper introduces a BM-based conditional training method for universal lesion detection in CT scans, addressing anchor imbalance and lesion shape diversity, leading to improved detection accuracy without extra annotation costs.

Contribution

It proposes a novel BM-based conditional training approach that enhances lesion detection by reducing anchor imbalance and adaptively modeling lesion sizes, outperforming existing methods.

Findings

Significant accuracy improvements across four state-of-the-art methods.

Effective handling of diverse lesion shapes and sizes.

No additional lesion mask annotations needed.

Abstract

Universal Lesion Detection (ULD) in computed tomography plays an essential role in computer-aided diagnosis. Promising ULD results have been reported by coarse-to-fine two-stage detection approaches, but such two-stage ULD methods still suffer from issues like imbalance of positive v.s. negative anchors during object proposal and insufficient supervision problem during localization regression and classification of the region of interest (RoI) proposals. While leveraging pseudo segmentation masks such as bounding map (BM) can reduce the above issues to some degree, it is still an open problem to effectively handle the diverse lesion shapes and sizes in ULD. In this paper, we propose a BM-based conditional training for two-stage ULD, which can (i) reduce positive vs. negative anchor imbalance via BM-based conditioning (BMC) mechanism for anchor sampling instead of traditional IoU-based rule; and (ii) adaptively compute size-adaptive BM (ABM) from lesion bounding box, which is used for improving lesion localization accuracy via ABMsupervised segmentation. Experiments with four state-of-the-art methods show that the proposed approach can bring an almost free detection accuracy improvement without requiring expensive lesion mask annotations.