An artificial intelligence (AI)-based tool designed to automate the delineation of proliferation centers (PCs) and provide an objective approach to chronic lymphocytic leukemia (CLL) was presented at the 2021 American Society of Hematology (ASH) Annual Meeting.

“Distinguishing CLL with many PCs from accelerated CLL (aCLL) or Richter transformation (RT) can be challenging, particularly in small needle-biopsy specimens. We manually annotated 25, 28, and 21 regions of interest (ROIs) encompassing small round PCs and confluent or expanded PCs of 10 CLL, 12 aCLL, and 8 RT digitized hematoxylin and eosin-stained slides, respectively,” explained Siba Hussein, MD, University of Rochester Medical Center, New York, and co-researchers.

All ROIs with both length and width larger than 2,000 pixels were analyzed. The tile length and stride were set to 1,000 and 100 pixels, respectively, with the ability to extract sufficient tiles from each ROI. The researchers quantified the nuclear size and intensity of cells occupying each tile to recreate PCs.

To note, nuclear size varied from 8 to 108 square micrometers, and nuclear mean intensity varied from 0 to 255.

“We generated heatmaps based on the heat values per tile inside each ROI from the 3 disease entities (CLL, aCLL, and RT). Areas with high heat values are shown in the yellow spectrum and correspond to tiles harboring cells with increased nuclear size and mean intensity (PCs in CLL cases and expanded/confluent PCs in aCLL and RT cases),” continued Dr Hussein and co-authors.

Low heat values appear in a blue spectrum in contrast to high heat values and correspond to tiles with decreased nuclear size and mean intensity (small neoplastic lymphocytes surrounding PCs). A heat value histogram was developed per tile for each ROI.  

“The 2 optimal thresholds isolated to obtain the highest separation value among the 3 disease entities based on the optimal F-score were: 0.228, below which the case was most likely CLL, and 0.288, above which the case was most likely RT. Cases with heat values ranging between 0.228 and 0.288 were most likely aCLL cases,” elaborated Dr Hussein and co-researchers.

The mean heat between the 3 entities showed a significant difference in the ranges of frequencies for CLL, aCLL, and RT, which were 0.168 to 0.233, 0.212 to 0.307, and 0.261 to 0.353, respectively.

“We describe a novel AI-based heatmap technique to objectively assess the extent of PCs in CLL, based on the integrative analysis of cell nuclear size and mean nuclear intensity. Using the mean heat value of all cases, we were able to reliably separate the three entities in question with robust diagnostic predictive values,” concluded Dr Hussein, et al.—Alexa Stoia

Hussein S, Chen P, Medeiros J, et al. Artificial Intelligence-Assisted Mapping of Proliferation Centers in Chronic Lymphocytic Leukemia/Small Lymphocytic Lymphoma Identifies Patterns That Reliably Distinguish Accelerated Phase and Large Cell Transformation. Presented at: the 2021 ASH Annual Meeting; Dec. 11-14; 2021; Abstract 1558.