Deep Learning for Real-Time Classification of...

Deep Learning for Real-Time Classification of Blood Cells in Microscopic Imaging

Author Name : Vinayak Swapnil Mokashi,Venkat Vedant Krishna, Rushi Patel, Riya Paunikar, Sandeep Vemulapalli, Naveena Gorrepati, Maremanda Venkata Srinivasa Saiteja, Ms. Trisha Hegde

ABSTRACT Blood tests play a critical role in medical diagnostics, with the identification and quantification of white blood cells (WBCs) serving as important indicators of a patient's health. Traditionally, laboratory technicians perform manual microscopic evaluations of WBCs, which is time-consuming and labor-intensive. Although specialized instruments for WBC segmentation and classification exist, they are often expensive and inaccessible for many doctors or healthcare facilities. To address this challenge, various computational techniques have emerged, offering improved efficiency and outcomes in recent years. This study focuses on the design and implementation of an Artificial Neural Network (ANN) for classifying different types of white blood cells from microscopic blood sample images. The research encompasses segmentation, feature extraction, and classification, aiming to categorize WBCs into five distinct types using the ANN model. Experimental analysis was conducted on microscopic images for the classification of WBCs. First, all images were segmented to isolate regions of interest. Next, feature vectors were extracted from these segmented images. The extracted features were then used as input for the ANN, enabling the classification of each image into its respective WBC type. Three feature sets were utilized to evaluate the classifier's performance. The segmentation process revealed that k-means clustering outperformed Otsu thresholding, with an average segmentation accuracy of 91.6% and 88.2%, respectively. Furthermore, the designed classifier achieved classification accuracy ranging from 93.8% to 96.5%, based on the features extracted from the segmented images. This research demonstrates that the use of ANN can significantly accelerate medical analysis, particularly as the volume of blood samples increases