AI Tools Enhance Detection of Immune Structures in Melanoma, Improving Survival Predictions

AI-Driven Detection of Tertiary Lymphoid Structures in Melanoma

Researchers from the ECOG-ACRIN Cancer Research Group have made a significant breakthrough in the field of melanoma research by applying artificial intelligence (AI) tools to detect tertiary lymphoid structures (TLS) in tumor tissue. This innovative approach has greatly enhanced the identification of TLS and improved survival predictions for patients with operable stage III/IV melanoma 1

1

2

The Importance of Tertiary Lymphoid Structures

TLS are key biomarkers associated with better prognosis and improved survival in melanoma patients. However, their detection is not yet a standard part of pathology reports, and manual identification is labor-intensive and prone to variability. The new AI-driven approach aims to standardize and streamline this process 1

1

Study Methodology and Findings

The research team, led by Dr. Ahmad A. Tarhini and Dr. Xuefeng Wang, conducted a retrospective analysis of thousands of archived digital images and corresponding RNA sequencing data from 376 patients with advanced, high-risk melanoma. These patients were part of the E1609 trial, which tested immune checkpoint blockade and cytokine therapy in high-risk melanoma 1

1

2

Key findings from the study include:

1. TLS were present in 55% of the E1609 cohort.
2. Patients with TLS showed significantly better overall survival (36.23% vs. 29.59% at 5 years).
3. Multiple TLS presence correlated with even better survival rates (38.04% for >1 TLS vs. 28.65%).
4. TLS density was also a significant prognostic factor for overall survival 1

   1

   2

   2

   .

AI Tools and Techniques Used

The researchers employed two main AI tools in their study:

1. HookNet-TLS: An open-source deep learning algorithm used to measure TLS and germinal centers (GC) within digitized H&E-stained slides. The model was retrained for improved accuracy 1

   1

   2

   2

   .

2. Gigapth Whole-Slide Foundation Model: This tool was used for digital pathology feature extraction and enhanced visualization of H&E image tiles through principal component analysis (PCA) 1

   1

   2

   2

   .

Related Stories

AI-powered atlas reveals how tertiary lymphoid structures predict cancer treatment outcomes

AI analyzes tumor biopsies to predict immunotherapy success in rare cancers, quadrupling survival

AI Models Show Promise in Predicting Triple-Negative Breast Cancer Prognosis

AI Tool SCORPIO Predicts Cancer Immunotherapy Response Using Routine Blood Tests

Implications and Future Directions

Dr. Tarhini emphasized the potential of this AI-driven approach to standardize TLS assessment using low-cost H&E-stained images. This could lead to improved prognostication and stratification within the American Joint Committee on Cancer (AJCC) staging system 1

1

Dr. Wang noted that the PCA visualizations generated by the Gigapth Foundation Model show promise in enhancing TLS and GC detection. Further fine-tuning of these results is ongoing 1

1

2

Potential Impact on Patient Care

The new survival prediction methods leverage low-cost, easily accessible technologies, potentially accelerating the adoption of TLS testing for high-risk melanoma patients. This could aid in discussions between patients and physicians regarding potential immunotherapy benefits 1

1

2

As this research continues to develop, it demonstrates the growing potential of AI tools in transforming cancer diagnosis, prognosis, and treatment planning. The integration of these technologies into clinical practice could lead to more personalized and effective care for melanoma patients in the future.