A new study conducted by Hospital for Special Surgery investigators in New York City has reported that their nuclei density is a robust tool for quantifying inflammatory burden in RA and it corroborates with multiple orthogonal measurements of inflammation. The present findings published in the journal ACR Open Rheumatology is a giant leap in automating the counting of individual nuclei in hematoxylin and eosin images. The algorithm developed by the researchers uses computer vision techniques including local adaptive thresholding, color deconvolution, watershed segmentation, and shape filtering.
Guan et al. have adapted computer vision algorithms to quantify nuclei density on synovial tissue obtained from arthroplasty samples. The nuclei density findings were subsequently compared with other measures used for quantifying RA inflammation namely clinical measures of disease activity, semiquantitative histology scores, and gene-expression data. A median of 112,657 (range 8,160-821,717) nuclei per synovial sample was detected using the algorithm. The corresponding sensitivity and specificity noted on the basis of pathologist-validated results were 97% and 100%.
The mean nuclei density estimated by the algorithm was found to be significantly high (P < 0.05) in synovium with elevated plasma cells, histology scores for lymphocytic inflammation, and lining hyperplasia. The findings of RNA sequencing analysis noted 915 differentially expressed genes, which correlated with nuclei density. Mean nuclei density was found to be significantly increased (P < 0.05) in subjected with higher levels of erythrocyte sedimentation rate, C-reactive protein, rheumatoid factor, and cyclized citrullinated protein antibody.
Presently, hematoxylin and eosin stain is the most commonly used technique for the histological assessments of synovial tissues. Considering the heterogeneity of synovial inflammation, variability in the chosen high-power fields for assessment can cause to difference in scores. The current study findings show that machine learning model helps in accentuating the research processes, thereby to optimize future patient care.
Reference: Guan S, Mehta B, Slater D, et al. Rheumatoid Arthritis Synovial Inflammation Quantification Using Computer Vision. ACR Open Rheumatol. 2022;4(4):322-331.