RFC-Net: Learning High Resolution Global Features for Medical Image Segmentation on a Computational Budget

TL;DR

RFC-Net is a novel neural network architecture that efficiently learns high-resolution global features for medical image segmentation, balancing accuracy and computational cost.

Contribution

Introduces RFC-Net with a Loose Dense Connection Strategy and m-way Tree structure to improve high-resolution feature learning efficiently.

Findings

Achieves state-of-the-art results on Polyp segmentation benchmarks.

Effectively balances high-resolution feature learning with reduced computational resources.

Demonstrates superior performance compared to existing methods.

Abstract

Learning High-Resolution representations is essential for semantic segmentation. Convolutional neural network (CNN)architectures with downstream and upstream propagation flow are popular for segmentation in medical diagnosis. However, due to performing spatial downsampling and upsampling in multiple stages, information loss is inexorable. On the contrary, connecting layers densely on high spatial resolution is computationally expensive. In this work, we devise a Loose Dense Connection Strategy to connect neurons in subsequent layers with reduced parameters. On top of that, using a m-way Tree structure for feature propagation we propose Receptive Field Chain Network (RFC-Net) that learns high resolution global features on a compressed computational space. Our experiments demonstrates that RFC-Net achieves state-of-the-art performance on Kvasir and CVC-ClinicDB benchmarks for Polyp segmentation.