Oops, you're using an old version of your browser so some of the features on this page may not be displaying properly.

MINIMAL Requirements: Google Chrome 24+Mozilla Firefox 20+Internet Explorer 11Opera 15–18Apple Safari 7SeaMonkey 2.15-2.23

Poster session 20

1434P - Machine learning features derived from spatial interaction of immune cell families are associated with survival in NSCLC patients post-immunotherapy

Date

21 Oct 2023

Session

Poster session 20

Topics

Clinical Research;  Statistics;  Immunotherapy

Tumour Site

Non-Small Cell Lung Cancer

Presenters

sara arabyarmohammadi

Citation

Annals of Oncology (2023) 34 (suppl_2): S755-S851. 10.1016/S0923-7534(23)01943-9

Authors

S. arabyarmohammadi1, G. Corredor1, M. Lopez de Rodas Gregorio2, K. Schalper3, A. Madabhushi1

Author affiliations

  • 1 Biomedical Engineering, Emory University, 30322 - Atlanta/US
  • 2 Pathology And Medical Oncology, Yale University School of Medicine, 06520 - New Heaven/US
  • 3 Pathology And Medical Oncology, Yale School of Medicine - Interventional Radiology Yale Cancer Center, CT 06510 - New Haven/US

Resources

Login to get immediate access to this content.

If you do not have an ESMO account, please create one for free.

Abstract 1434P

Background

The spatial architecture of B and T cells in the tumor microenvironment (TME) is known to be associated with treatment response in cancer patients. However, more complex spatial statistics to characterize the interaction of more than two immune cell types could further inform our understanding of tumor response to immunotherapy (IO). In this work we sought to evaluate quantitative metrics of higher order spatial relationships between multiple (>2) cell types to better characterize and predict overall survival (OS) in NSCLC patients.

Methods

The study included quantitative immunofluorescence images of pre-IO tumor biopsy specimen of 135 NSCLC patients from five centers with OS information. To predict OS, a set of 223 image features that characterize the spatial interplay of different cell types (CD4+, CD20+, CD8+, tumor, and stroma) in the TME were extracted. First, every nucleus belonging to each immune cell family was detected using an image analysis program (MATLAB) to identify the cell’s expression, subsequently the cell coordinates were recorded. The algorithm then identifies each triplet of distinct cells in proximity of each other to form spatial triangles. Features relating to the density, neighborhood, and proximity of nuclei were subsequently extracted. A risk score (RS) was then computed for each patient using the least absolute shrinkage and selection operator. To predict OS, the Cox regression model was trained on 100 patients (St) and validated on 35 patients (Sv). The discriminative ability of the features was assessed using Kaplan-Meier and log-rank tests.

Results

The RS was significantly associated with OS in both St (HR: 1.76; 95% CI = 1.09 - 2.85; P = .02), and Sv (HR: 2.8; 95% CI= 1.18–6.68; P=.01). RS based risk stratification showed greater accumulation of tumor, CD4 T cells, and CD20 B cells in very close proximity of each other in low-risk patients, compared to high-risk patients.

Conclusions

Machine learning derived features describing the higher order spatial relationships between tumor, CD20 B cells, and CD4 T cells were found to be associated with OS in NSCLC patients treated with IO. Additional, independent multi-site validation of these findings is warranted.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

K. Schalper: Financial Interests, Speaker, Consultant, Advisor: Agenus, Ariagen, AstraZeneca, Bristol Myers Squibb , Clinica Alemana de Santiago, Genmab, Merck Sharp and Dohme, OnCusp Therapeutics, Repertoir Therapeutics, Takeda Oncology; Financial Interests, Institutional, Research Grant: Akoya Biosciences, AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb , Eli Lilly and Company , Merck Sharp and Dohme, Navigate Biopharma, Pierre-Fabre, Ribon Therapeutics, Surface Oncology, Takeda Oncology, Tesaro, Inc.; Financial Interests, Advisory Board: EMD Serono, Shattuck Labs; Financial Interests, Research Grant: Moderna Inc. A. Madabhushi: Financial Interests, Personal, Advisory Board, Serve on SAB and consult: SimbioSys; Financial Interests, Personal, Advisory Board: Aiforia, Picture Health; Financial Interests, Personal, Full or part-time Employment: Picture Health; Financial Interests, Personal, Ownership Interest: Picture Health, Elucid Bioimaging, Inspirata Inc; Financial Interests, Personal, Royalties: Picture Health, Elucid Bioimaging; Financial Interests, Institutional, Funding: AstraZeneca, Bristol Myers Squibb, Boehringer Ingelheim, Eli Lilly. All other authors have declared no conflicts of interest.

This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.

For more detailed information on the cookies we use, please check our Privacy Policy.

Customise settings
  • Necessary cookies enable core functionality. The website cannot function properly without these cookies, and you can only disable them by changing your browser preferences.