GGB Color Normalization and Faster-RCNN Techniques for Malaria Parasite Detection

Malaria is a disease transmitted by a female mosquito anopheles bite. Malaria commonly occurs in tropical and sub-tropical regions that having minimum health facilities. Promising news for us, early malaria diagnosis is a proven effective preventive a malaria-related mortality. In addition, automated malaria detection studies have shown a promising performance in reducing the manual microscopy-based examination times. However, since the quality input image is not standardized, a proper image preprocessing technique is notable in recognizing the object. Therefore, this study applies green, green, blue (GGB) color normalization as a preprocessing step in malaria detection. We evaluate our technique in a large public dataset containing 2,418 infected thin blood smear images by 49,900 parasites. The results show that our technique has malaria detection performance consistently better sensitivity and consistently similar precision in several intersection over union (IoU) thresholds. Furthermore, it indicates that using GGB color normalization in malaria parasite detection is valuable in reducing the false positive error.