Chapter 6 - Application of machine learning and image processing for detection of breast cancer

Abstract

Information technology is playing an important role in healthcare systems. Communication technologies enable the user to stay in contact with the physician. The physician is also staying in contact with his patient. Smart wearable sensing devices are easily available in a smart environment. The patient wears these sensing devices and performs daily life routine actions. If there is any disturbance in the patient body than sensing devices sense it and send the message to the doctor about the patient condition. Through this process, the physician can give proper treatment to a patient according to the condition of the patient at that time. Patient life can be saved using these innovations in smart healthcare systems. Breast cancer is a critical healthcare problem. The computer can solve this problem using advanced algorithms of artificial intelligence (AI). Breast cancer is the second leading disease causes woman deaths. According to a report, 2.5 million breast cancer cases identified in the United States in 2017. According to IARC, 8.2 million deaths caused by cancer per year. Deaths rate from cancer is increasing developed countries every year. Cancer is the name of uncontrolled cell growth in the body. Breast cancer occurs due growth of cells in the breast. Group of extra tissues is known as tumor. Cancer is the deadliest disease in the whole world. Cancer has stages. It can be cured in the first stage. After the first stage, it can never be cured. But in the first stage of cancer, it is very difficult to diagnose it. Early stage detection of cancer can save millions of lives. Computer-aided detection is used for early detection of cancer. X-ray image of the breast is known as a mammogram. Breast is compressed between the two plates and the x-ray beam is applied to take mammogram. The technique which is applied before symptoms occur in woman breast is known as screening mammography. Mammography images have noise. Image processing techniques are used to remove noise from a mammogram. In the first-phase preprocessing is done. Filters are applied to remove noise from mammograms. After preprocessing noise-free image is passed through the segmentation phase. In segmentation, the region of interest (ROI) is found. ROI is the region in the image in which a tumor exists. After the segmentation feature is extracted. The whole image’s dataset is passed from above phases. After this, text or comma-separated values (CSV) file is obtained which has the features of mammogram dataset. Dataset has two labels “0” and “1.” Label “1” indicates cancer and label “0” indicates noncancer image. Feature extracted dataset is also labeled using mammogram labeled. Subfield of AI is machine learning (ML). It has two major types that are supervised and unsupervised. Supervised learning has labeled data for training. Unsupervised learning is trained using unlabeled data. Dataset is divided into two parts. One is used for training is known as training dataset. Second is used for testing is known as test data. Training data are used for training and testing data are used for the testing model. In the end, a different measure is calculated using different formulas such as precision, accuracy, and f-score using confusing matrix. If the reasonable accuracy is achieved when the model is considered for prediction. Feature is extracted from new samples and label is predicted for that case. Either case has cancer or not. Our work is beneficial to save human lives. We applied ML algorithms to detect cancer from a mammogram. We achieved reasonable accuracy for the prediction of cancer.