Effectiveness of using computer aided detection based on convolutional neural network for screening microcalcification On USG Mammae

Abstract

Breast cancer is a very common cancer in women, affects millions of women each year, and is also one of the leading causes of cancer deaths among women. Early-stage cancer detection can reduce breast cancer mortality significantly in the long term. Breast ultrasound is a cost-effective and widely available screening tool. Breast ultrasound can be used for women who are at high risk of developing breast cancer but cannot undergo MRI examinations or for pregnant women who should not be exposed to X-rays, and also for women who have dense breast tissue. Although ultrasound is considered to be used for breast screening, its use is highly dependent on the doctor’s experience in reading the images. Therefore, to diagnose breast cancer using ultrasound images automatically can use computer assistance, namely Computer-Aided Detection (CADe). This study discusses a deep learning-based CADe system, namely the convolutional neural network (CNN), CNN is desired to evaluate several ultrasound images to make a microcalcification diagnosis. Microcalcifications are fine white spots, similar to grains of salt. They are not usually cancer, but certain patterns can be an early sign of cancer. Detecting microcalcification using CNN can be used as a routine breast screening, which can help doctors find signs of breast cancer earlier than is currently possible. The best MSE(Mean Squared error) loss value obtained is achieved at 0.24 with an accuracy of 0.76.