Classification of Thalassemia Based on Red Blood Cell Images using Deep Learning

The advancements in artificial intelligence (AI) have proven to be highly beneficial across various domains, particularly in the field of image processing. One significant application of AI-driven image processing is in the medical field, where it facilitates early disease detection. Thalassemia, a genetic blood disorder, is one such condition that can be predicted with the assistance of AI. This study employs deep learning techniques to detect Thalassemia based on red blood smear imaging. The images are classified into three classes Thalassemia, Iron Deficiency Anemia (IDA), and Normal Red Blood Cells (RBC) images. Deep learning pretrained network AlexNet and ResNet-50 were utilized for image classification. The experimental results demonstrate that both models perform satisfactorily, with accuracy rates surpassing 95%. ResNet-50 exhibited superior overall performance, achieving an average accuracy of 99.36% ± 1.62%, along with a precision of 98.62% ± 2.86%, sensitivity of 99.74% ± 0.45%, specificity of 98.01% ± 4.44%, and an F-score of 98.49% ± 3.56%. Despite ResNet-50's higher accuracy, AlexNet demonstrated greater efficiency in training time, completing the process in an average of 270 seconds. These findings highlight the potential of deep learning models, particularly ResNet-50, in enhancing the early detection of Thalassemia.