Rethinking the transfer learning for FCN based polyp segmentation in colonoscopy

TL;DR

This paper introduces a novel training scheme for FCN-based polyp segmentation in colonoscopy images, combining segmentation and classification tasks to reduce overfitting and improve accuracy on challenging datasets.

Contribution

It proposes an interactive weight transfer mechanism between dense and coarse vision tasks, enhancing segmentation performance and robustness against small, imbalanced datasets.

Findings

Achieved 4.34% and 5.70% improvements in Polyp-IoU on two datasets.

Demonstrated effectiveness of combined segmentation and classification training.

Reduced overfitting in deep neural networks for polyp segmentation.

Abstract

Besides the complex nature of colonoscopy frames with intrinsic frame formation artefacts such as light reflections and the diversity of polyp types/shapes, the publicly available polyp segmentation training datasets are limited, small and imbalanced. In this case, the automated polyp segmentation using a deep neural network remains an open challenge due to the overfitting of training on small datasets. We proposed a simple yet effective polyp segmentation pipeline that couples the segmentation (FCN) and classification (CNN) tasks. We find the effectiveness of interactive weight transfer between dense and coarse vision tasks that mitigates the overfitting in learning. And It motivates us to design a new training scheme within our segmentation pipeline. Our method is evaluated on CVC-EndoSceneStill and Kvasir-SEG datasets. It achieves 4.34% and 5.70% Polyp-IoU improvements compared to the state-of-the-art methods on the EndoSceneStill and Kvasir-SEG datasets, respectively.