Computational-Efficient Morphology Zone search in Peripheral Blood Smear for Identification of Atypical Blood Cell Morphologies
Yogeswarr Sathish, Sree Niranjanaa Bose, Aarthi Sathya Narayanan, Shivani Parayil, Bharathi Kannan Natarajan, Varalakshmi Perumal, S Hemanth, Murali Mohan · 2025
Advances in Digital Pathology has created milestones in creating value add to the pathologists fraternity. One of such solution that aids in reducing the workload of the hematopathologists is the automated analysis of the Peripheral Blood Smear (PBS). The current systems work more towards intermediate magnification to achieve significant success in the Turn-Around-Time (TAT), but same is not achieved for higher magnification. This paper describes above edge-compute integrating neural network model that works with higher magnified smear images to identify the ideal zone of morphology. The results shows that the system is able to reach the region of interest with overall accuracy of 74% for 5-region and 90.3% for 3-region classification system. Further, the tested model achieves the TAT of less than 180 secs in higher magnification. The proposed system shall be implemented for computational constrained imaging systems that may be restricted to use single objectives.