A Study on Blood-Cell Segmentation Method for the Identification of Hematological Disorders

A important task in the detection of hematological abnormalities is the automated segmentation of blood cells. It is essential for diagnosis, arranging treatments, and assessing results. This procedure uses a hybrid blood-cell segmentation technique based on RESNET50 Unet that may be utilized to identify a number of hematological diseases. Our main contributions are a more precise seed-point and better segmentation performance achieved by combining RESNET50 Unet techniques while keeping the advantages of both methods. It is a computationally effective strategy since it combines algebraic and non-iterative geometric algorithms. The minor and major axes should also be estimated using the residue and residue offset factors, according to our proposal. The residue offset parameter that is here presented results in better segmentation with appropriate EF. Modern approaches are contrasted with our method. It performs better than the current EF methods in terms of precision, Jaccard score, and F1 score as well as dice similarity. Other medical and cybernetics applications could benefit from it. Our suggested model beat current models on the test set, with an average accuracy of 97.5%. Also, we contrasted the performance of our model with that of other segmentation models, including DeepLabv3+, Mask R-CNN, and U-Net. Our ResNet50-based model outperformed these models in terms of accuracy and speed, according to the results. As a result, our suggested strategy utilising ResNet50 is a potential technique for precise and effective blood cell segmentation, which can help in the early identification of blood-related disorders