Abstract 6176: Identifying poor prognosis stage I lung adenocarcinoma patients through novel morphological biomarker based on computational pathology
Bibliographic record
Abstract
Abstract Current practice for evaluating prognosis of resectable lung adenocarcinoma (LUAD) patients relies on the Tumor, Node, Metastasis staging system. Despite low relapse risk, post-recurrence survival in a stage I population is poor with a median post relapse survival of around 25 months. There is a need for a biomarker that can enrich for stage I patients with relapse risk to identify a population likely to benefit from adjuvant therapy. In this study, we aim to identify a predictive biomarker of relapse in early-stage lung cancer following surgical resection using AI-based computational pathology. The retrospective patient cohort analyzed in this study consisted of a total of 166 patients that underwent surgical resection for a clinical stage I LUAD. Out of these, 54 patients experienced disease recurrence within 5 years, and 112 patients had a confirmed disease-free period of at least 5 years post-surgery. For each patient, at least one digitized slide (average = 2.46) stained with haematoxylin and eosin (H&E) was available for analysis. The project was accepted by the IUCPQ ethics committee (2022-3751, 22138). Manual annotations delineating the tumor core (TC) were drawn by pathologists. We then applied several proprietary image analysis models capable of distinguishing epithelium, stroma, and necrosis within the tumor tissue as well as detecting tumor infiltrating lymphocytes (TILs) and cell nuclei. In this project, we additionally introduce a novel approach rooted in mathematical graph theory to encoding topological aspects of the tissue morphology. A cell-graph constructed from the positions of cell nuclei was used to calculate cell-level graph-theoretic measures which were then aggregated to the slide-level by computing the median. In total, 23 data readouts were obtained of which 14 were based on the standard image analysis pipeline alone, and a further 9 originated from our newly developed graph-based approach. We identified promising biomarkers using the Wilcoxon rank-sum test and the area under the ROC curve (AUROC); robustness of this evaluation was investigating using resampling methods, using 100 bootstrap samples and 100 repeats of 3-fold cross-validation, respectively. This allowed us to evaluate the prognostic value of the biomarkers with respect to 5-year relapse or death. We were able to identify the best prognostic biomarker for relapse to be the median weighted clustering coefficient (CC) across all cells in the TC area, which is derived from the graph-based analyses. The median CC was able to achieve a whole-cohort AUROC of 0.695 (average cross-validated AUROC = 0.675, 95% interval 61.7%-70.4%) as well as a relapse vs relapse-free Wilcoxon test p-value of 4.65x10-5 (bootstrapped average = 0.0002). In addition, we were also able to demonstrate that the biomarker shows some relation to the architecture patterns. Citation Format: Florian J. Song, Alma Andoni, Manal Kordahi, Sara Batelli, Armin Meier, Markus Schick, Emilie Mahieu, Günter Schmidt, Claire E. Myers, Michael Abadier, Abjihit Dasgupta, Christopher Abbosh, Darren Hodgson, Michèle Orain, Fabien C. Lamaze, Yohan Bossé, Philippe Joubert. Identifying poor prognosis stage I lung adenocarcinoma patients through novel morphological biomarker based on computational pathology [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6176.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".