Mammography Image Quality Assurance Using Deep Learning.

Tobias Kretz,Clemens Elster,Klaus-Robert Mueller,Tobias Schaeffter

doi:10.1109/tbme.2020.2983539

Abstract

According to the European Reference Organization for Quality Assured Breast Cancer Screening and Diagnostic Services (EUREF) image quality in mammography is assessed by recording and analyzing a set of images of the CDMAM phantom. The EUREF procedure applies an automated analysis combining image registration, signal detection and nonlinear fitting. We present a proof of concept for an end-to-end deep learning framework that assesses image quality on the basis of single images as an alternative. Virtual mammography is used to generate a database with known ground truth for training a regression convolutional neural net (CNN). Training is carried out by continuously extending the training data and applying transfer learning. The trained net is shown to correctly predict the image quality of simulated and real images. Specifically, image quality predictions on the basis of single images are of similar quality as those obtained by applying the EUREF procedure with 16 images. Our results suggest that the trained CNN generalizes well. Mammography image quality assessment can benefit from the proposed deep learning approach. Deep learning avoids cumbersome pre-processing and allows mammography image quality to be estimated reliably using single images.

Highlights

M AMMOGRAPHY screening using x-ray radiation is an important diagnostic tool and routinely applied for early detection of breast cancer [1], [2]
We provide a detailed comparison of the deep learning approach with the EUREF Guideline procedure based on simulated and real images of the contrast-detail phantom for mammography (CDMAM) phantom
Our results demonstrate that a trained convolutional neural net (CNN) can be used to successfully analyze images of the CDMAM phantom in mammography image quality assurance

Summary

Introduction

M AMMOGRAPHY screening using x-ray radiation is an important diagnostic tool and routinely applied for early detection of breast cancer [1], [2]. Since cancerous tissue is denser than healthy tissue, its x-ray attenuation is slightly higher, which results in a contrast difference in the image acquired. The EUREF Guideline analysis results in a contrast-detail curve [6] that characterizes the ability to detect small structures of the phantom in dependence on their size. A contrast-detail curve consists of twelve points assigned to images of the CDMAM phantom [7]. Each point consists of a diameter and the minimum thickness needed such that a signal

Objectives

Methods

Results

Discussion

Conclusion