A model for mammogram image segmentation based on hybrid enhancement

Abstract

Introduction: Detection of breast cancer at early stages is very important and it helps in saving the life of the patient. The mammography is one of the techniques which is used worldwide by radiologists to detect breast cancer. It is very important to detect breast cancer correctly as human life is associated with correct detection. So, nowadays computer-aided diagnose (CAD) systems are used to detect breast cancer from a mammography image. In this work, a transfer learning-based method is used to train the system for the detection of breast cancer. The proposed model is based on a hybrid technique of image enhancement and segmentation for the pre-processing. The pre-processing is done to improve the peak signal to noise ratio (PSNR) value of originally acquired images. The Mammography Image Analysis Society (MIAS) dataset is used in this work along with other clinical images which result in the creation of around 928 images in the dataset. Aim: This work aims to segment the affected area image from the original image. Then enhanced the segmented image by using PSNR so that better result is helpful to identify the affected area. The correct detection is associated with a life decision. Result: Mammogram image has been segmented from the affected area. Various type of noises may be induced in the black and white image. The image represents the original, enhanced and segmented image. The PSNR value for an image varies from 33 to 36. The segmented image is representing the affected regions to limit the image and to eliminate useless details. Conclusion: This paper put forward the methodology which will segment the affected area from the image. The image has been enhanced based on LCM and CLAHE. Morphology technique has been used along with the Otsu threshold for accurate segmentation. The proposed network is promising as it is trained for mammogram dataset for higher accuracy.