Keywords: lesion, segmentation, detection, CNN, UNet
TL;DR: With the right reweighing strategy, small pathology detection performance can be improved while maintaining segmentation accuracy.
Abstract: There are many clinical contexts which require accurate detection and segmentation of all focal pathologies (e.g. lesions, tumours) in patient images. In cases where there are a mix of small and large lesions, standard binary cross entropy loss will result in better segmentation of large lesions at the expense of missing small ones. Adjusting the operating point to accurately detect all lesions generally leads to oversegmentation of large lesions. In this work, we propose a novel reweighing strategy to eliminate this performance gap, increasing small pathology detection performance while maintaining segmentation accuracy. We show that our reweighing strategy vastly outperforms competing strategies based on experiments on a large scale, multi-scanner, multi-center dataset of Multiple Sclerosis patient images.
Paper Type: methodological development
Primary Subject Area: Segmentation
Secondary Subject Area: Application: Radiology
Paper Status: original work, not submitted yet
Source Code Url: Source code for the loss reweighing module and associated metrics will be made publically available here: https://github.com/brennanNichyporuk/MIDL-2021-Short-Paper-Code. Our full codebase contains unpublished work, but will be made publically available if we are invited to submit an extended version of our work in MedAI.
Data Set Url: Our MRI data was obtained over the course of a privately funded clinical trail. We do not have permission from the company in question to release the data publically.
Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.
Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.
Source Latex: zip