398P - AI-powered whole-slide image analysis of tumor-infiltrating lymphocytes for prediction of prognosis in colorectal cancer

Background

The degree of T-cell infiltration has been suggested as an important prognostic biomarker for colorectal cancer (CRC) patients, regardless of other clinical and/or pathological factors. In this study, we analyzed tumor-infiltrating lymphocyte (TIL) counts of CRC using Lunit SCOPE IO, an artificial intelligence (AI)-powered whole slide image (WSI) software analyzer. Our aim was to analyze the prognostic significance of AI-powered TIL analysis in CRC.

Methods

Lunit SCOPE IO was trained and validated with a 2.8 x 10⁹ micrometer² area and 5.9 x 10⁶ TILs from 3,166 H&E Whole-Slide Images (WSI) of multiple cancer types, annotated by 52 board-certified pathologists. The Inflamed Score (IS) was defined as the proportion of all tumor-containing 1 mm²-size tiles within a WSI classified as being of the inflamed immune phenotype (high TIL density within cancer epithelium). H&E images, sequencing data and survival data of stage I-III CRC patients from The Cancer Genome Atlas (TCGA) were utilized for this analysis.

Results

Stage I-III CRC samples (n = 461) with clinical data were analyzed. The median of IS was 8.56 (IQR 3.74-18.39). IS showed moderate positive correlations with CD8A (rs = 0.422, p < 0.001) and CD3G (rs = 0.377, p < 0.001) expression levels but weaker positive correlations with regulatory T cells (rs = 0.162, p < 0.001), TH1 (rs = 0.209, p < 0.001) or TH2 cell proportions (rs = 0.128, p = 0.006). The IS was higher in CMS1 group compared to CMS 2-4 groups (median 18.49 vs. 6.90, p < 0.001). No significant differences in IS was observed across TNM stages. The recurrence-free survival of the patients with IS higher than third quartile (>= 18.39) were significantly longer compared to the lower group (p = 0.034, HR 0.540, 95% CI 0.306-0.954). The same outcome was observed in cases with MSS tumors (p = 0.023, HR 0.380, 95% CI 0.165-0.877).

Conclusions

AI-powered analysis of WSI can provide prognostic information in stage I-III CRC patients. Further development of AI-powered TIL analysis including the spatial information of TIL in relation to tumor cells may improve prognostic power.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Lunit Inc.

Disclosure

Y. Lim: S. Song; S. Ahn; J. Ryu; H. Song; M. Ma; S. Park; S. Pereira; B.J. Aum; S. Shin; S. Cho; K. Paeng; D. Yoo; W. Jung; C. Ock: Financial Interests, Personal, Full or part-time Employment: Lunit Inc. All other authors have declared no conflicts of interest.