Main Article Content
Acute Myeloid Leukemia (AML) is a form of the condition that is fatal and has a high mortality rate. It is characterised by abnormal cells growing rapidly inside the human body. The conventional method for detecting AML seems to be examining the blood sample manually under a microscope, which is a manual and cumbersome task that also requires well-trained medical expertise for efficient identification. On the other hand, considering medical diagnosis, the capacity to classify medical images faster and accurate is essential. The classification of medical images my currently be accomplished using a range of methodologies including Machine Learning (ML), Deep Learning (DL) and Transfer Learning (RL). While these approaches are effective for large datasets, they can take a while and~not ideal for small datasets. In recent years, advances in Deep Convolutional Neural Networks (DCNN) have made it possible and produce more accurate and promising outcome while processing a~medical image. However, the paradigm that DCNN~use for training includes a large number of annotations in order to prevent overfitting and produce promising results. Obtaining large-scale semantic annotations in clinical operations might be problematic in some cases, particularly biological expertise knowledge is needed. It is also regular occurrence in scenarios where only a small number of annotated classes are accessible in some circumstances. At this context, in order overcome the drawback of traditional approach a framework has been developed which comprises of Enhanced Few-Shot Learning Technique integrated Base Classifier (Feature Encoder)-EFLTBC. The proposed model has built using base classifier and meta-learning block, and it optimized the better results. To diagnose AML, the doctor must count the number of white blood cells and red blood cells and see if there are any abnormal health conditions in that using a microscope. However, obtaining an accurate result takes time and effort. To address these issues, the proposed Novel AML detection model employing is used in this study. Base classifier utilizing ResNet-18 pretrained model and meta learning block has computed using the average feature of every samples. Also, the dataset that we used consisting of three classes includes Normal monocytes, Abnormal monocytes, Lymphocyte and Experimental results outperform various existing deep learning technique with the accuracy of 97%, recall of 96.55% F1-Score of 96.65% and precision of 96.60.