Segmentation and classification of lymphoblastic leukaemia using quantum neural network

Abstract

AbstractIn lymphoblastic leukaemia (ALL), the bone marrow naturally produces immature cells. Each year ALL is diagnosed with over 6500 instances, and the trend is still going upward. Technological advancements in AI and big data analytics help doctors and radiologists make accurate and efficient clinical decisions. The proposed method consists of two core steps: segmentation and classification based on the quantum convolutional networks. A three‐dimensional U‐network is proposed having 70 layers that are trained on the optimal hyperparameters, which provides 0.98 dice scores. The four‐qubit quantum transfer learning model is proposed for classifying different types of blood cells. The accuracies achieved are 0.99 on blast cells, 0.99 on Basophils, 0.98 on Eosinophils, 0.97 on Neutrophils, 0.99 on Lymphocytes, and 0.96 on Monocytes. The proposed classification model provides 0.99 average accuracy.