Segmentation of Microscope Erythrocyte Images by CNN-Enhanced Algorithms

Area of Science:

• Medical Imaging
• Computational Biology
• Image Processing

Background:

• Accurate segmentation of erythrocytes (red blood cells) is crucial for quantitative analysis in hematology.
• Manual analysis of erythrocyte images is time-consuming and prone to subjectivity.
• Automating erythrocyte segmentation and analysis can improve diagnostic efficiency and accuracy.

Purpose of the Study:

• To develop and validate a robust two-stage algorithm for segmenting and analyzing erythrocyte images.
• To compute key statistical parameters of erythrocytes, including count, size, and shape ratios.
• To enhance the precision of erythrocyte segmentation through a multi-stage processing approach.

Main Methods:

• Image preprocessing included noise reduction using median, mean, and bilateral filters.
• Background subtraction was performed using a rolling ball filter.
• Segmentation combined distance transform, Otsu's method, and watershed algorithm, followed by morphological operations and iterative refinement.

Main Results:

• The two-stage segmentation algorithm significantly improved precision from 0.857 to 0.968 in artificial image tests.
• The algorithm successfully computed statistical object values and performance metrics like sensitivity and specificity.
• The method demonstrated potential for classifying segmented erythrocytes, identifying abnormal or poorly segmented cells.

Conclusions:

• The proposed two-stage segmentation algorithm offers a significant improvement in the accuracy and precision of erythrocyte image analysis.
• This automated approach has the potential to replace manual methods, reducing subjectivity and increasing efficiency.
• The technique is adaptable for deep learning pipelines, serving as a probability map processor.