Inception V3 Based Transfer Learning Model for the Prognosis of Acute Lymphoblastic Leukemia from Microscopic Images

Acute lymphoblastic leukemia (ALL) cancer is a serious, rapidly growing disease that can be lethal. It is one of the most common pediatric cancer all over the world. Even the slightest delay to identify ALL, can put one's life at stake. Diagnosing leukemia traditionally requires many lab tests and a physician's prognosis, both of which can be time intensive and erroneous. The task of distinguishing lymphoblasts cells from normal cells can be difficult since the two cells look identical morphologically under the microscope. In this study, an efficient and effective transfer-learning-based CNN model was proposed which could precisely classify the affected cell from the data distribution. The proposed model was based on Inception V3 architecture. This architecture was selected after comparing the performances of other state-of-the-art convolutional neural network models. Inception V3 architecture was used for feature extraction in the first stage and then trainable dense layers with various numbers of neurons are added further for deep neural connection and classification. This part was introduced to find the necessary features and map them to get better classification results. The suggested model achieved overall accuracy of 98.00%, precision of 98.00% and F1-score of 98.00%. The suggested model may aid our physicians to diagnose ALL precisely and rapidly.