Do Noises Bother Human and Neural Networks In the Same Way? A Medical Image Analysis Perspective

TL;DR

This paper investigates whether noise affects human and neural network perception similarly in medical image analysis, proposing a framework optimized for neural network denoising that outperforms human-vision-based methods.

Contribution

It introduces an application-guided denoising framework tailored for neural networks, diverging from traditional human-vision-centric approaches, and demonstrates superior results.

Findings

Proposed framework outperforms human-vision denoising methods.

Framework effective across multiple datasets and models.

Neural network-focused denoising enhances medical image analysis.

Abstract

Deep learning had already demonstrated its power in medical images, including denoising, classification, segmentation, etc. All these applications are proposed to automatically analyze medical images beforehand, which brings more information to radiologists during clinical assessment for accuracy improvement. Recently, many medical denoising methods had shown their significant artifact reduction result and noise removal both quantitatively and qualitatively. However, those existing methods are developed around human-vision, i.e., they are designed to minimize the noise effect that can be perceived by human eyes. In this paper, we introduce an application-guided denoising framework, which focuses on denoising for the following neural networks. In our experiments, we apply the proposed framework to different datasets, models, and use cases. Experimental results show that our proposed framework can achieve a better result than human-vision denoising network.