ODRNN: Optimized Deep Recurrent Neural Networks for Automatic Detection of Leukaemia

Abstract

Abstract Leukaemia, a kind of cancer that may occur in individuals of all ages, including kids and adults, is a significant contributor to worldwide death rates. This illness is currently diagnosed by manual evaluation of blood samples obtained using microscopic imaging, which is frequently slower, lengthy, imprecise. Additionally, inspection under a microscope, leukemic cells look and develop similarly to normal cells, making identification more difficult. Convolutional Neural Networks (CNN) for Deep Learning has provided cutting-edge techniques for picture classification challenges throughout the previous several decades, there is still potential for development with regard to performance, effectiveness, and learning technique. As a consequence, the study provided a unique deep learning approach known as Optimized Deep Recurrent Neural Network (ODRNN) for identifying Leukaemia sickness by analysing microscopic images of blood samples. Deep recurrent neural networks (DRNN) are used in the recommended strategy for diagnosing Leukaemia, then the Red Deer Optimization algorithm (RDOA) applies to optimize the weight gained by DRNN. The mass of DRNN from RDOA will be tuned on the deer roaring rate behavior. The model that has been proposed is evaluated on two openly accessible Leukaemia blood sample datasets, AML, ALL_IDB1 and ALL_IDB2. It is possible to create an accurate computer-aided diagnosis for Leukaemia malignancy by using the proposed deep learning model, which shows encouraging results. The research work uses statistical metrics related to disease including specificity, recall, accuracy, precision and F1 score to assess the effectiveness of the proposed model for identification and classification. The proposed method achieves highly impressive results, with scores of 98.96%, 99.85%, 99.98%, 99.23%, and 99.98%, respectively.