Enhancing Blood Platelet Counting through Deep Learning Models for Advanced Diagnostics

Platelet counting is considered an essential factor for the diagnosis of blood clotting disorders as well as other illnesses as COVID-19 and Leukemia. Hemocytometer-based manual techniques are labor-intensive and prone to errors, especially in dynamic settings where platelet clumping develops over time. This work presents a novel method based on a collection of regression and object detection models- MobileNetv2, EfficientDetD0, and SSD MobileNet-that were trained on a large number of microscopic blood smear images. Multiple models have been trained and analysed to find out which models perform well for our dataset and problem. This study demonstrates how deep learning and object detection in blood platelet counting can revolutionize the field by increasing diagnostic precision and efficiency in healthcare systems. Out of all the models trained, we observed that EfficientDet D0 gave us the best performance, with a precision of 0.64, mAP of 0.96 and recall of 0.65.