Machine learning approach for leukemia detection

Abstract

Blood cancer(Leukemia) represents a critical global health burden. Early detection and prompt intervention by medical health providers are critical to improving medical outcomes. In our research we use Convolutional Neural Networks (CNN), and other traditional methods like RF-Random Forest (RF), neighboring based method KNN, classification method SVM to predict potential cancerous cells and classify them into benign and malignant cells. The training dataset undergoes preprocessing methods. It improves performance and feature extraction in models. Among the four different approaches used CNN shows higher accuracy of 99.97% for benign cells and 100% for malignant subtypes. For evaluation of the models performance we utilize accuracy, precision, recall and F1-score metrics. Experimental results from our research highlighted the advantage of deep learning, i.e CNN is found to be the best approach in Image classification. The user-friendly GUI for input and prediction display enhances usability. The findings form our research states the need of AI driven diagnostic tools in early blood cancer detection. Further work will be focused on larger datasets and real-world clinical validation.