Revolutionizing Malaria Diagnosis: Precision Segmentation of Thin Blood Smears via Advanced U-Net Convolutional Architectures
Muhammad Shameem P, Muthukumaran Malarvel · 2024
The most recent advances in deep learning have demonstrated considerable promise for medical image analysis, especially for the segmentation of thin blood smears, which are microscopic and used to diagnose malaria. This paper presents an improved Convolutional Neural Network (CNN) model for the segmentation of malaria parasites in thin blood smear images, which is based on the U-Net architecture. Our model aims to address the problems that are frequently present in microscopic images, such as overlapping cells and variable stain intensity. Specifically, our model is quite good at overcoming the inherent difficulties in microscopy imaging, especially when it comes to segmentation in conditions with overlapping cells, poor staining quality, and low contrast. Our model stands out for its ability to handle these complex scenarios, which is a major advancement over traditional segmentation strategies. The U-Net architecture is a perfect fit for this purpose because of its efficiency in segmenting biomedical images. We provide an extensive analysis of the model's performance, contrasting it with conventional segmentation techniques in terms of accuracy, sensitivity, and specificity.