AI based morpho-class net classifier with HOG approach for bone marrow cell classification

A crucial stage in haematological diagnostics is the classification of bone marrow cells, and the combination of artificial intelligence and image processing offers a reliable, effective, and precise alternative for manual inspection. An automated image analysis system for precise categorization and detection in biomedical imaging is proposed in this work. In order to reduce noise while maintaining crucial structural information, the system initially acquires input images and then pre-processes them using a bilateral filter. The Otsu thresholding technique is then used to segment the pre-processed image, separating the region of interest from the background for better analysis. Following segmentation, key texture and shape information from the segmented regions is captured through feature extraction utilizing the Histogram of Oriented Gradients (HOG) approach. An artificial intelligence model based AI Morpho-ClassNet is then fed these extracted features in order to classify the photos into various diagnostic categories. In order to facilitate additional interpretation, the system ultimately produces an output image that highlights the categorized regions. The proposed approach shows potential for precise, effective, and automated image-based diagnosis and analysis across a range of biological applications by combining noise reduction, segmentation, feature extraction, and deep learning-based classification into a single process. The proposed system is implemented in python and it achieves the accuracy of 97.5%.