Acute lymphoblastic leukemia detection using ensemble features from multiple deep CNN models

We presented a methodology for detecting acute lymphoblastic leukemia (ALL) based on image data. The approach involves two stages: Feature extraction and classification. Three state-ofthe-art transfer learning models, InceptionResnetV2, Densenet121, and VGG16, were utilized to extract features from the images. The extracted features were then processed through a Global Average Pooling layer and concatenated into a flattened tensor. A linear support vector machine (SVM) classifier was trained and tested on the resulting feature set. Performance evaluation was conducted using metrics such as precision, accuracy, recall, and F-measure. The experimental results demonstrated the efficacy of the proposed approach, with the highest accuracy achieved at 91.63% when merging features from VGG16, InceptionResNetV2, and DenseNet121. We contributed to the field by offering a robust methodology for accurate classification and highlighted the potential of transfer learning models in medical image analysis. The findings provided valuable insights for developing automated systems for the early detection and diagnosis of leukemia. Future research can explore the application of this approach to larger datasets and extend it to other types of cancer classification tasks.