Computer-assisted quantification of CD3+ T cells in follicular lymphoma

Area of Science:

• Digital pathology
• Computational pathology
• Immunohistochemistry analysis

Background:

• High-resolution digital scans of pathology slides enable computer-based image analysis for immunohistochemistry (IHC) quantification.
• Current computer-based methods need refinement for routine clinical use, requiring evaluation in clinical settings.

Purpose of the Study:

• To pilot a computerized cell quantification method for estimating CD3 positive (CD3+) T cells in follicular lymphoma (FL).
• To assess the comparability of computerized quantification to expert pathologist estimations.

Main Methods:

• A computerized method using entropy-based histogram thresholding for CD3+ cell segmentation after color space transformation.
• Evaluation by four expert hematopathologists using visual estimation and manual cell marking on 20 FL images.
• Calculation of sensitivity and specificity for computer segmentation against pathologist readings.

Main Results:

• The computerized method demonstrated high agreement with manual cell marking, surpassing visual estimations.
• Mean sensitivity and specificity of computer segmentation were 90.97% and 88.38%, respectively.
• High correlation coefficients (Lin's 0.81, Spearman's 0.96) between computerized quantification and pathologist readings, exceeding inter-pathologist agreement.

Conclusions:

• Computerized quantification of CD3+ T cells in FL shows strong agreement with expert pathologist assessments.
• The developed algorithm offers a reliable alternative to manual methods, with potential for integration into clinical workflows.
• Future applications include investigating the relationship between tumor architecture and patient outcomes.