An Open-Source, Automated Tumor-Infiltrating Lymphocyte Algorithm for Prognosis in Triple-Negative Breast Cancer

Abstract

Purpose: Although tumor-infiltrating lymphocytes (TIL) assessment has been acknowledged to have both prognostic and predictive importance in triple-negative breast cancer (TNBC), it is subject to inter and intraobserver variability that has prevented widespread adoption. Here we constructed a machine-learning based breast cancer TIL scoring approach and validated its prognostic potential in multiple TNBC cohorts. Experimental Design: Using the QuPath open-source software, we built a neural-network classifier for tumor cells, lymphocytes, fibroblasts, and “other” cells on hematoxylin–eosin (H&E)–stained sections. We analyzed the classifier-derived TIL measurements with five unique constructed TIL variables. A retrospective collection of 171 TNBC cases was used as the discovery set to identify the optimal association of machine-read TIL variables with patient outcome. For validation, we evaluated a retrospective collection of 749 TNBC patients comprised of four independent validation subsets. Results: We found that all five machine TIL variables had significant prognostic association with outcomes (P ≤ 0.01 for all comparisons) but showed cell-specific variation in validation sets. Cox regression analysis demonstrated that all five TIL variables were independently associated with improved overall survival after adjusting for clinicopathologic factors including stage, age, and histologic grade (P ≤ 0.0003 for all analyses). Conclusions: Neural net-driven cell classifier-defined TIL variables were robust and independent prognostic factors in several independent validation cohorts of TNBC patients. These objective, open-source TIL variables are freely available to download and can now be considered for testing in a prospective setting to assess clinical utility. See related commentary by Symmans, p. 5446

Objective assessment of tumor infiltrating lymphocytes as a prognostic marker in melanoma using machine learning algorithms

“…The machine-defined TILs variables were constructed using five different methods, as previously described. 17 The first and established method was to calculate eTIL% representing the proportion of TILs over tumor cells, calculated as (TILs/TILs + tumor cells) x 100. 1 Four additional methods were used to measure TILs as follows: Measurement of the proportion of TILs over total cells: etTIL % = (TILs/total cells) x 100 Measurement of the proportion of TILs over stromal cells: esTIL % = (TILs/total cells – tumor cells) × 100 Measurement of the density of TILs over tumor region: eaTILs (mm 2 ) = TILs/sum of tumor region areas analysed (mm 2 ) Measurement of the density of TILs over stromal area: easTIL % = [sum of TIL area (mm 2 ) /stroma area (sum of tumor region areas analysed (mm 2 ) – sum of tumor cell area (mm 2 ))] × 100 …”

Artificial intelligence in breast cancer histopathology

“…TIL detection has been framed as a nuclear classification and segmentation problem. Earlier nuclear segmentations had been performed using thresholding, 47 watershed algorithms, 48 and many variations of active contours. 49,50 Although performance is limited by variations in staining, overlapping nuclei, and small and complex shapes of nuclei, these algorithms remain popular in research.…”

Dissecting Tumor-Immune Microenvironment in Breast Cancer at a Spatial and Multiplex Resolution

“…Recently, recommendations have been published from the international immuno-oncology group to aid towards clinical validation [103]. In BC, the clinical significance of digital TILs enumeration has been explored [65,[104][105][106][107]. In a recent study by Bai et al, a semisupervised neural network was developed for the scoring of digital TILs in 920 early TNBC patients and demonstrated that increasing digital TILs density was a favorable prognostic factor [104].…”

“…In BC, the clinical significance of digital TILs enumeration has been explored [65,[104][105][106][107]. In a recent study by Bai et al, a semisupervised neural network was developed for the scoring of digital TILs in 920 early TNBC patients and demonstrated that increasing digital TILs density was a favorable prognostic factor [104]. Furthermore, retrospective analyses of prospective cohorts that recruited a large sample of BC patients, concluded that digital TILs counts independently predicted pCR [105,106].…”

“…However, since H&E manual TILs scoring remains the "gold standard", the concordance of digital and manual evaluation needs to be further assessed. Indeed, significant correlations between the algorithm's TILs scores and the pathologist's read sTILs assessments have been reported [64,104]. Intriguingly, multiplex methods can also digitally assess lymphocytic subpopulations densities and the interplay between TILs and CAFs in H&E-stained images has been also studied with the use of digital pathology.…”