Computer Vision-Based Microcalcification Detection in Digital Mammograms Using Fully Connected Depthwise Separable Convolutional Neural Network.

Khalil Ur Rehman,Jianqiang Li,Tariq Mahmood,Yan Pei,Anaa Yasin,Saqib Ali

doi:10.3390/s21144854

Abstract

Microcalcification clusters in mammograms are one of the major signs of breast cancer. However, the detection of microcalcifications from mammograms is a challenging task for radiologists due to their tiny size and scattered location inside a denser breast composition. Automatic CAD systems need to predict breast cancer at the early stages to support clinical work. The intercluster gap, noise between individual MCs, and individual object’s location can affect the classification performance, which may reduce the true-positive rate. In this study, we propose a computer-vision-based FC-DSCNN CAD system for the detection of microcalcification clusters from mammograms and classification into malignant and benign classes. The computer vision method automatically controls the noise and background color contrast and directly detects the MC object from mammograms, which increases the classification performance of the neural network. The breast cancer classification framework has four steps: image preprocessing and augmentation, RGB to grayscale channel transformation, microcalcification region segmentation, and MC ROI classification using FC-DSCNN to predict malignant and benign cases. The proposed method was evaluated on 3568 DDSM and 2885 PINUM mammogram images with automatic feature extraction, obtaining a score of 0.97 with a 2.35 and 0.99 true-positive ratio with 2.45 false positives per image, respectively. Experimental results demonstrated that the performance of the proposed method remains higher than the traditional and previous approaches.

Highlights

Breast cancer is the leading cancer resulting in death among women around the globe.The mortality rates are very high in Asia and Africa compared to Europe according to theWorld Health Organization International Agency for Research on Cancer (WHO-IRAC) [1].The automatic detection of suspected lesions from mammograms at the early stages could help radiologists in the prediction of breast cancer in less time to avoid unnecessary biopsies in order to reduce mortality rates [2]
This study proposes a novel diagnostic method based on computer vision and a fully connected, depthwise-separable convolutional neural network (FC-DSCNN) to detect microcalcification clusters and classify these clusters to predict breast cancer at the early stages
We use the traditional deep 2D convolutional neural networks (CNNs) for comparing our proposed method on the same procedure; In the diagnosis phase, the fully automated computer-vision-based FC-DSCNN predicts breast cancer from digital mammograms; Our proposed method can achieve a better sensitivity compared to previous studies. The evaluation metrics such as the recall, F1-score, and area under the curve (AUC) curve are calculated to verify the model’s diagnostic ability; in this study, we evaluated for the first time the local PINUM dataset and the public DDSM dataset for the prediction of breast cancer with our proposed approach and achieved a higher true-positive rate; The evaluation of the local dataset could help doctors and radiologists diagnose cases of breast cancer in women at the initial stages in real time

Summary

Introduction

The automatic detection of suspected lesions from mammograms at the early stages could help radiologists in the prediction of breast cancer in less time to avoid unnecessary biopsies in order to reduce mortality rates [2]. Mammography is a standard screening method for the detection of breast cancer, but it is not a golden standard for breast cancer diagnostics. Mammography is an effective tool for the detection of breast cancer in women at the early stages. Microcalcification develops in the breast and is one of the main signs of breast cancer visible on mammograms at the early stages. Microcalcifications are tiny tissues having a diameter from 0.1 mm to 0.7 mm that comprise features such as shape, morphology, and location, which can be extracted in the phase of preprocessing to differentiate malignant from benign lesions [3]. Microcalcifications are more likely to be benign if the diameter is 0.1 mm or under 0.5 mm, while a heterogeneous diameter smaller than 0.5 mm is malignant [4]

Objectives

Results

Conclusion