Acute myelogenous leukemia detection using optimal neural network based on fractional black-widow model

Identifying the hematological structure of leukocytes or white blood corpuscles is often very complex in medical practices. The higher mortality rate associated with the number of new cases of Acute Myelogenous Leukemia (AML) is increased day by day due to improper treatment and delay in diagnosis. Recently, the hematological experts used microscopic analysis of peripheral blood smears to diagnose hematological disorders. Nevertheless, these kinds of detection processes are more time-consuming, complex, and costly with inaccurate results. Hence, we proposed a Fractional Black Widow-based Neural Network (FBW-NN) for the detection of AML. The Adaptive Fuzzy Entropy (AFE) Model has been introduced to segment the AML region. AFE is the amalgamation of both the active contour-based model and the fuzzy C-mean clustering method. The statistical and image-level features are extracted after the segmentation process. Subsequently, the Fractional Black Widow Optimization is formulated to enhance the performance of the Artificial Neural Network (ANN). A Fractional Black Widow-based Neural Network is proposed to detect the AML. The proposed work is implemented in the MATLAB platform with different performance metrics. Experimentally, the proposed method accomplishes better detection rates than other state-of-the-art methods such as pre-trained DCNN, Naive Bayes, CNN, and CSC-ACNN. When compared to these methods, the proposed FBW-NN yields 96.56% accuracy, 97.81% specificity, 96.90% sensitivity, 97.20% precision, and 97.90% recall results.

Automated summary

The study proposed a detection method based on Fractional Black Widow Neural Network to improve the detection performance of AML. Experimentally, the method achieved better accuracy, specificity, sensitivity, precision, and recall compared to other state-of-the-art methods.