Fuzzy-CL: Fuzzy Rank-Based Ensembling Aided Contrastive Learning for Malaria Detection Using Red Blood Cell Smears
Shreyan Kundu, Rahul Talukdar, Semanti Das, Souradeep Mukhopadhyay, Soumalya Mallick, Biswadip Basu Mallik, Swarnamouli Majumdar · 2025
Many academics have worked to identify malaria-parasitized cells in blood sample images using Deep Learning algorithms throughout the years. Even though malaria is quite harmful, it can be controlled if caught early enough. This provides motivation to put in place a precise malaria detection method that can take the place of the current labor-intensive manual procedure. The manual procedure entails counting the red blood cells that are parasitized and those that are not, as well as visually inspecting the blood samples. This is a labor-intensive procedure that can be botched by an untrained medical staff member and takes a long time. Keeping these factors in mind, our goal was to create a solution that would require less training for medical personnel to utilize, saving them time and labor. After reading through a number of research studies on the application of deep learning methods to malaria detection, we have developed a model that fills in the gaps in existing systems without sacrificing the precision of the findings. With a 97.61% accuracy rate, our suggested model—which combines fuzzy ranking based ensembling with triplet loss aided contrastive learning (CL)—performs better than several state-of-the-art models.