Hematological Malignancy Identification via K-means based ROI Extraction

• Dr. Latha Kiran Krishna Rajendran

  Author

Identifying hematological diseases early on and classifying them helps give patients improved clinical options.  Moreover, it is of great difficulty to perform the analysis of the image of a smear of peripheral blood. In addition, the analysis takes more time. Generally, an expert pathologist has a fairly heavy study load, having to study large numbers of images, and suffers a delay because of inter and intra-observer variability. Consequently, a number of CAD schemes were developed that automatically diagnose hematology diseases. As a result, laboratory technicians and pathologists may devote their time to more complicated activities. Please consult the internet for more information. This paper describes the automated detection of blood disorders using image processing with a computer-aided diagnostic system (CAD).  The technique implements various image pre-processing operations to reduce noise and improve contrast and morphological enhancement of the image for better segmentation. Moreover, we also use K-means clustering to extract the Region of Interest(ROI). In essence, the algorithm could identify leukocyte areas from the blood smear image. Geometrical, statistical, and textural feature descriptors, calculated from the extracted ROI, will describe the patterns of abnormal cells. The feature vectors are used by the classifiers, kNN and Naive Bayes, to classify malignant and non-malignant blood cells. The constructed post-concomitant CDSS produces a classification accuracy of 92.8% when applied.

• PDF

Copyright (c) 2026 International Journal of Clinical Research and Medical Sciences

This work is licensed under a Creative Commons Attribution 4.0 International License.