Artificial Intelligence Brings TILs Into the Digital Era

By Armen Gevorgyan, MD, Medical Oncologist, Editor In Chief of OncoDaily Breast, OncoDaily Lung Aug 26, 2026

Fact checked by Susanna Mikayelyan, MD, Managing Editor of OncoDaily Hematology Aug 26, 2026

Medically reviewed by Amalya Sargsyan, MD, Medical Oncologist, Vice President of Research and Intelligence at OncoDaily Aug 26, 2026

Tumor-infiltrating lymphocytes (TILs) have become one of the most important immune biomarkers in breast cancer, particularly in triple-negative breast cancer (TNBC), where the interaction between the immune system and tumor biology strongly influences clinical outcomes.

For years, TIL assessment has relied on expert pathologist evaluation of hematoxylin and eosin (H&E)-stained tumor sections. Although standardized approaches have improved reproducibility, manual scoring remains dependent on specialist expertise, requires significant time, and may be difficult to implement consistently across healthcare systems.

A new study published in The Lancet Oncology evaluated whether artificial intelligence (AI)-based computational tools could provide a reproducible alternative for measuring TILs and predicting outcomes in patients with early-stage TNBC.

The independent validation study, known as CATALINA, assessed previously developed AI algorithms for computational tumor-infiltrating lymphocyte (cTIL) quantification using large clinical datasets derived from randomized clinical trials. The findings provide important evidence supporting the role of digital pathology in immune biomarker assessment.

TILs

Why Tumor-Infiltrating Lymphocytes Matter in Triple-Negative Breast Cancer

Triple-negative breast cancer represents one of the most biologically aggressive breast cancer subtypes, characterized by the absence of estrogen receptor, progesterone receptor, and HER2 expression.

Unlike hormone receptor-positive breast cancer, where endocrine therapy provides a major therapeutic advantage, TNBC has historically relied heavily on chemotherapy. However, the recognition that many TNBC tumors contain active immune infiltration has changed the understanding of this disease.

TILs reflect the presence of immune cells within the tumor microenvironment and provide information about the interaction between the host immune system and malignant cells.

Previous studies have demonstrated that higher levels of stromal TILs are associated with improved prognosis in TNBC, including better invasive disease-free survival, distant disease-free survival, and overall survival.

The clinical importance of immune biomarkers has increased further with the introduction of immune checkpoint inhibitors, where identifying patients with immune-active tumors may help refine treatment strategies.

However, translating TIL assessment into routine clinical practice requires methods that are accurate, reproducible, and scalable.

The CATALINA Study: Independent Validation of AI-Based TIL Assessment

The CATALINA study was designed to determine whether AI-derived TIL scores could provide reliable prognostic information compared with traditional pathologist-based evaluation.

Researchers analyzed data from 1,759 patients, including 1,356 patients with complete clinicopathological information, pathologist-scored stromal TILs, and AI-derived cTIL scores available. The dataset included patients with early-stage triple-negative or HER2-positive breast cancer, with outcome analysis focused primarily on early TNBC cohorts.

Two previously validated AI pipelines were independently deployed without retraining or modification. The algorithms generated computational TIL measurements from digitized H&E whole-slide images and were evaluated against clinical outcomes from seven prospective randomized adjuvant breast cancer trials.

This approach was particularly important because independent external validation is required before AI-based biomarkers can be considered for clinical implementation.

AI-Derived TIL Scores Show Prognostic Value

The study demonstrated that AI-derived TIL measurements provided statistically significant prognostic information.

The researchers observed a moderate correlation between computational TIL scores and traditional pathologist-scored stromal TIL measurements, with correlation coefficients ranging from 0.375 to 0.473.

Both traditional stromal TIL assessment and AI-derived TIL scores were independently associated with improved outcomes, including:

invasive disease-free survival
distant disease-free survival
overall survival

After adjustment for clinical and pathological factors, higher TIL levels were associated with reduced risk of recurrence and death.

For invasive disease-free survival, the hazard ratio associated with pathologist-scored stromal TILs was 0.73, while AI-derived percentage lymphocyte scores showed a hazard ratio of 0.80. Similar associations were observed for distant disease-free survival and overall survival.

These results suggest that AI-generated immune measurements can capture meaningful biological information from routine pathology images.

AI Does Not Replace Pathologists, It Expands Access to Immune Biomarkers

Although AI-based scoring demonstrated prognostic value, the study also highlighted an important limitation.

When AI-derived TIL scores were combined with traditional clinicopathological factors and pathologist-based stromal TIL assessment, the additional prognostic contribution of AI measurements was no longer statistically significant.

This suggests that AI currently complements rather than replaces expert pathological evaluation.

The strongest clinical value of AI may come from settings where standardized TIL assessment is difficult to perform due to limited pathology resources, differences in expertise, or high clinical workload.

Digital pathology platforms could eventually allow consistent immune biomarker assessment across institutions and support broader implementation of precision oncology approaches.

The Role of Digital Pathology in Future Breast Cancer Treatment

The emergence of AI-based pathology represents a major shift in oncology diagnostics.

Traditional pathology relies on human interpretation of complex tissue patterns. AI algorithms can analyze thousands of cellular features, quantify immune infiltration, and identify patterns that may not be easily measurable through conventional approaches.

For breast cancer, this technology may become increasingly important as treatment decisions become more dependent on biological characteristics rather than anatomical staging alone.

Future applications may include:

improved risk stratification in early TNBC
integration with genomic and immune biomarkers
identification of patients most likely to benefit from immunotherapy
standardized assessment across global healthcare systems

However, prospective studies are still required to determine how AI-based TIL measurements should be incorporated into treatment algorithms.

Challenges Before Clinical Implementation

Despite promising results, several challenges remain before AI-based TIL scoring becomes routine clinical practice.

First, algorithms must demonstrate consistent performance across different scanners, staining protocols, institutions, and patient populations.

Second, regulatory frameworks must define how AI-generated pathology biomarkers should be validated and integrated into clinical workflows.

Finally, clinicians must determine whether AI-derived measurements provide meaningful clinical advantages beyond existing pathological assessment.

The CATALINA study represents an important step because it evaluated AI tools in a large independent dataset rather than only in development cohorts.

The Future: From Visual Pathology to Quantitative Immune Profiling

The future of breast cancer pathology is moving toward quantitative, data-driven immune profiling.

TILs have already established themselves as an important biomarker in TNBC. AI technology now provides a potential pathway to make immune assessment faster, more reproducible, and more accessible.

The CATALINA validation study demonstrates that computational TIL scoring can provide clinically relevant prognostic information, supporting the continued development of AI-powered biomarkers in breast oncology.

As immunotherapy and precision medicine continue to evolve, combining digital pathology with molecular and clinical data may help create a more complete understanding of each patient’s tumor biology.

The Bottom Line

Artificial intelligence-based tumor-infiltrating lymphocyte quantification represents an important advance in breast cancer biomarker development.

In the CATALINA independent validation study, AI-derived TIL scores demonstrated prognostic value in early triple-negative breast cancer and improved risk discrimination compared with traditional clinical and pathological factors alone.

Although AI does not replace expert pathology assessment, it may expand access to standardized immune biomarker evaluation and accelerate the transition toward more personalized breast cancer care.

The next generation of breast cancer diagnostics may not only identify what a tumor looks like, but also quantify how the immune system interacts with it.

Author

Armen Gevorgyan

MD, Medical Oncologist, Editor In Chief of OncoDaily Breast, OncoDaily Lung

Editor In Chief of OncoDaily Breast, OncoDaily Lung Armen Gevorgyan, MD, is a medical oncologist at Mikayelyan University Hospital and Editor-in-Chief of OncoDaily Breast and Lung. His professional interests focus on thoracic oncology and breast cancer, with a particular emphasis on advancing clinical practice, oncology education, and scientific communication. Through his clinical and editorial work, Dr. Gevorgyan contributes to the dissemination of evidence-based oncology knowledge and the development of global oncology dialogue.

Fact checked by Susanna Mikayelyan MD, Managing Editor of OncoDaily Hematology Susanna Mikayelyan, MD, is a hematology resident physician at the Yeolyan Hematology and Oncology Center and a medical writer with the OncoDaily Intelligence Unit. Her medical education began at Yerevan State Basic Medical College, where an interest in chemistry and pharmacology led her to study pharmacy from 2016 to 2019. During her final year, she was awarded a nominal scholarship and graduated with honors. She subsequently entered the Faculty of General Medicine at Yerevan State Medical University. While studying medicine, a growing interest in disease and therapeutic mechanisms led her to pursue a subspecialty deeply grounded in pathophysiology and pharmacology. She graduated from Yerevan State Medical University with honors in 2025 and, later that year, entered residency training in hematology, with a particular clinical interest in hematologic oncology. Since April 2026, Susanna has been a medical writer with the OncoDaily Intelligence Unit. Her work focuses on hematologic malignancies, spanning disease biology, emerging research, clinical trials, and scientific meetings, synthesizing complex evidence into clear, clinically relevant content for physicians, researchers, and the wider oncology community. Outside medicine, she enjoys art history, cinematography, football, motorsports, and learning foreign languages through literature and self-study.

Medically reviewed by Amalya Sargsyan MD, Medical Oncologist, Vice President of Research and Intelligence at OncoDaily Amalya Sargsyan, MD, MSc, is a medical oncologist in Yerevan, Armenia, and Vice President of Research & Intelligence at OncoDaily. She heads the Sarcoma Service at D'Clinic, treats adult solid tumors at the Adult Solid Tumors and Chemotherapy Clinic of the Yeolyan Hematology and Oncology Center, and leads the Adult Solid Tumor Team at the Immune Oncology Research Institute. Her clinical practice covers sarcoma, gastrointestinal cancers, and adolescent and young adult (AYA) oncology. She earned her MD and completed medical oncology residency at Yerevan State Medical University, then an MSc in Precision Medicine in Clinical Practice at the University of Cyprus. Her sarcoma training began at the Bank of Cyprus Oncology Centre and continued through a three-month fellowship at the Sarcoma Unit of Fondazione IRCCS Istituto Nazionale dei Tumori in Milan, organized with the European School of Oncology, followed by observerships at Memorial Sloan Kettering Cancer Center and the sarcoma program at Stanford Medicine. She trained in gastrointestinal oncology under the mentorship of Yelena Janjigian at MSK, as a recipient of the ASCO Conquer Cancer International Development and Education Award and Memorial Sloan Kettering GI Oncology International Training Award. Her research addresses access and equity in cancer care in low- and middle-income countries. She is principal investigator of the IMMONKG study, a multinational retrospective cohort examining alternative immune checkpoint inhibitor dosing strategies across LMICs, and first author of the JCO Global Oncology analysis of immunotherapy access in Armenia's out-of-pocket health system (Sargsyan et al., 2025). She has authored and contributed to peer-reviewed publications in journals including Nature Reviews Clinical Oncology, JCO Global Oncology, The Lancet Oncology, and Expert Review of Gastroenterology & Hepatology. She has received ESMO Leadership and Career Development Award in 2026, the ESMO Merit Award twice and the ASCO Conquer Cancer International Development and Education Award. At OncoDaily she directs the Research & Intelligence unit, overseeing global oncology content strategy, editorial operations across six disease verticals, and more than 50 scientific events a year - including the How I Treat virtual summit series. She is an Adjunct Assistant Professor at Yerevan State Medical University, founder of the Young Oncology Group of Armenia, and founder of the ASCO Oncology Student Interest Group at Yeolyan.