A Deep-Learning AI Framework to Diagnose Acute Lymphoblastic Leukemia
Victoria Chong
Eric Hamber Secondary
Floor Location : S 068 N

Acute Lymphoblastic Leukemia (ALL) is the most common pediatric cancer worldwide. It’s malignant and can be fatal within as little as a few weeks, if not treated as soon as possible. It’s crucial to detect in its early stages when survival rates are at their highest. Previously, feature classification and engineering was applied, but is a time-consuming and difficult process. Deep learning, on the other hand, is capable of feature extraction and classification and is therefore, a revolutionary technology. The global shortage of physicians in rural areas combined with the shortage of equipment, causes patients of malignancy to suffer the consequences, namely late and misdiagnosis. This really challenging problem is exactly what I set out to solve through this application. I developed an Image Classification System, using MATLAB and Alexnet, which can detect the presence of leukemia from basic microscopic blood samples. By inputting a large dataset of blood samples, which are malignant or non-malignant, the algorithm was able to extract features and later detect the presence of leukemia. With this technology, people who have ALL, with no expertise, can get a predicted probability of malignancy. By attaching a foldscope (paper microscope) to the back of any smartphone, with a blood sample slide in it, the algorithm can analyze the sample and predict an accurate and timely diagnosis.