Authors

Sara Noor Eldin, Jana Khaled, Ganna Tamer, Maysoon Hossam, Dr. Ammar Mohamed, Eng. Noha Elmasry, and Eng. Menna Gamil

Publishing Date

14/11/2020

Abstract

According to the world health organization, Cancer is the second leading cause of death globally. The most common cancer among women is Breast Cancer. Detecting the type of Breast Cancer that a patient has accurately is crucial. An accurate diagnosis ensures that the patient is given the correct and the most effective treatment. Manual Histopathology exams done by pathologists are tiresome, takes a lot of time. The results might differ from one pathologist to another due to the difference in expertise, and other human errors. So, the main idea of our proposed system is to detect breast cancer types from microscopic Biopsy images using Deep Learning algorithms. The proposed system will determine whether the suspicious cells in these images are benign or malignant. If it is malignant, then it is either in situ or invasive carcinoma. We aim to decrease the time it takes for the results to reach the worried and anxious patients, as well as cutting down any chance of a misdiagnosis that might lead to other serious complications.

1.1 Background

Breast cancer is a type of cancer where there is an abnormal and uncontrollable growth in some cells crowding out the breast normal cells. Though breast cancer is most commonly thought of as a terrible disease that only affects women. For men, the lifetime risk of getting breast cancer is about 1 in 833, so there are uncommon but serious cases of male breast cancer. There are many types of breast cancer, and correctly identifying each one is important to determine the proper treatment. They can be divided into two overarching groups, the carcinomas, and the sarcomas. Carcinomas are cancers that arise from the epithelial component of the breast that consists of the cells that line the lobules and terminal ducts. Sarcomas are rare cancers that arise from the Stromal (connective tissue) components of the breast, which account for less than 1% of primary breast cancers. There are many subdivisions in the Carcinomas. The first major division is between in situ which refers to cancer in which abnormal cells have not spread beyond where they first formed, and invasive carcinoma which means that the cancer has “invaded” or spread to the surrounding breast tissues . While their are other special types like colloid (mucinous), medullary, micropapillary, papillary, and tubular. So it is really important to distinguish between these various sub-types of Breast Cancers because they can have different prognoses and treatment implications. Breast Cancer is detected and diagnosed through tests varying from mammograms, breast ultrasounds, MRIs, and breast biopsy. A doctor might recommend a biopsy if he or she finds something suspicious during a physical exam or one of the other mentioned tests. A biopsy is the main way doctors diagnose most types of cancer. Other tests can suggest that cancer is present, but only a biopsy can make a diagnosis . A biopsy is the surgical removal of a representative sample of tissue from a suspicious lesion. Histopathology is then performed by pathologists where they examine the sample of tissues that undergo a process called ‘Histology’ under the microscope to study the cell’s architecture and patterns so they could make the diagnosis. Biopsy also has other types like the Fine Needle Aspiration (FNA), Core Needle Biopsy, CT-guided biopsy, and ultrasound guided biopsy . This manual identification of cancer from microscopic biopsy images is subjective in nature and may vary from a pathologist to another depending on their expertise and other factors . As well as the long duration that this process takes to make a certain diagnosis. So, we propose a system to automate this delicate procedure for the pathologists by identifying and classifying cancerous cells from microscopic biopsy images using Deep Learning Algorithms.

1.2 Motivation

Breast cancer has been known for a very long time, dates back to 3,000–2,500 B.C.E. Cancer has been growing ever since. Breast Cancer is the most common cancer for women in Egypt up to 38.8%. The causes of breast cancer are not yet discovered, however, family genetics has a great role in the percentage of being affected by the disease. Not only family history is a cause but also age is a very big risk for cancer development, most women who are diagnosed with cancer their age ranges from 50 to 70 years old. Breast cancer checkups can help in reducing mortality. Detection of the cancer type is really important for the effectiveness of treating and managing the disease. Therefore Our work is motivated by the accuracy of pathology and the speed of deep learning combined together, getting outstanding results in detecting the type of cancer. Breast cancer is a disease that any woman might have the misfortune of being diagnosed with or know someone who did, we have conducted a questionnaire asking people who had done a breast cancer test or know someone who did some questions about their experience throughout the process.

1.3 Problem Statement

Due to variance in expertise and other factors, the breast cancer histopathology exams might be interpreted differently by pathologists leading to a misdiagnosis and the selection of a wrong treatment path. The challenge is automating the pathologist process in detecting the breast cancer type (Benign, Carcinoma, Situ, and Invasive Carcinoma) from microscopic Biopsy images, and decreasing the duration of the diagnosing process of results.