Automated Grid Detection in Hemocytometer Images Using Canny and Hough Line Transform for Cell Quantification

Accurate cell quantification using a hemocytometer requires identifying a defined Area of Interest (AOI) rather than analyzing the entire image. Manual selection of this region is time-consuming and error-prone, particularly when processing large datasets. To address this, we propose a fully automated method for AOI detection in hemocytometer images (microscope images of hemocytometer grids), aimed at improving lung cancer cell quantification. The original image is first converted to grayscale, then the Sobel algorithm is applied to generate two variants highlighting edge features in the horizontal and vertical directions, respectively. These variants are subsequently processed using the Canny edge detection algorithm and the Hough line transform to extract grid-like structures. However, this process often introduces noise, such as overlapping or redundant lines and irrelevant detections. To overcome this, we propose a filtering and identification algorithm to refine the results and accurately determine the signature lines that define the hemocytometer AOI. Experimental results confirm that the proposed method offers both high accuracy and processing speed compared to manual approaches. The fully automated system, which integrates AOI detection with a YOLOv8m-based cell detection model, achieved a precision of 0.91, recall of 0.87, and mAP@50 of 0.93. The overall F1-score reached approximately 0.891, reflecting a strong balance between sensitivity and specificity. In high-density cell samples, the average counting error compared to manual methods was below 11.3%, while in low-density samples, it was under 2.6%. These findings demonstrate the method's effectiveness and potential for reliable, standardized cell quantification in biomedical applications.