Acute Myeloid Leukemia Classification in High Resolution Blood Smear Image using Hybrid Attention-Based Deep Convolutional Belief Network - 21/11/26
, Keerthana Ganeshan 2, Sandeep Chandrika Sudhakaran 3, Premkumar Chinnammam Chettiyar 4, Kannammal Nagarajan 5Abstract |
Objectives |
In this research, the AML disease is classified using the Dual Attention-based Deep Convolutional Belief Network (DA-DConvBNet) model that helps in accurately classify the types of AML using blood smear images.
Patients and Methods |
The quality of the blood smear image is increased by enhancing the image’s pixels and eliminating the noise of the image through the Fast Non-Local Means (FNLM) Denoising method. The computation time of the DA-DConvBNet model is decreased by the utilization of a binary threshold-based Region of Interest (ROI) extraction method, which ignores the irrelevant regions and extracts the important regions. The most appropriate features for the accurate classification are extracted by the Hybrid Pre-trained Gray level Structural (HPGS) based features, which also facilitates minimizing the dimensionality, thus the risk of overfitting is eliminated. Overall, the DA-DConvBNet model is the better solution for accurate AML classification.
Results |
The DA-DConvBNet model achieved a sensitivity of 98.41%, accuracy of 98.40%, and specificity of 98.38% for the training data of 90% using the AML-Cytomorphology_MLL_Helmholtz dataset, respectively.
Conclusion |
The DA-DConvBNet model is developed in this research to classify the subtypes of AML accurately using the blood smear images.
Le texte complet de cet article est disponible en PDF.Keywords : Deep Learning, Acute Myeloid Leukemia, Dual Attention, Haematological disorder, Blood Smear Images
Plan
Bienvenue sur EM-consulte, la référence des professionnels de santé.
L’accès au texte intégral de cet article nécessite un abonnement.
Déjà abonné à cette revue ?
