Detection of Lymphoma from the Bone Marrow Microscopic Images using Convolutional Neural Networks

Cancer is defined by the growth of anomalous cell growth that divides rapidly and has the ability to infect and destroy normal bodily tissue. It is a fatal disease affecting people of all age groups. However, thanks to advancements in cancer detection, therapy, and prevention, survival rates for many forms of cancer are advancing. Lymphoma is one of the most common and fast progressive type of cancer that affects people of all ages. It starts in the immune system's infection-fighting cells, known as lymphocytes. These cells can be found in the lymph nodes, spleen, thymus, bone marrow, and other organs. Lymphocytes alter and develop uncontrollably in lymphoma patients. Follicular Lymphoma and Mantle Cell Lym phoma affects the older age group. Each year millions are diagnosed with this disease. One of the most important aspects for detecting the type of cancer is the blood count. Conventionally manual counting is done and yields a desirable result, but it consumes more time for processing. Deep learning approaches can address these issues by extracting beneficial properties from a large number of raw datasets in a short span of time. Since lymphoma requires early diagnosis, deep learning techniques can help save lives. The objective of the project is to detect if the patient has lymphoma using bone marrow microscopic images and to categorize the lymphoma into three classes namely- Follicular Lymphoma (FL) and Mantle Cell Lymphoma (MCL) and normal tissues using CNN architecture (Inception V3) and (ResNet-50) and to predict the accuracy and efficiency and compare with existing models. Image processing techniques such as segmentation, filters, have also been employed to locate the cancerous parts in the images.