Academic Journal
P13.12.A SUBTYPES AND SURVIVAL ANALYSIS ANALYSIS OF PRIMARY CENTRAL NERVOUS SYSTEM LYMPHOMA WITH RADIOMICS FEATURES
| العنوان: | P13.12.A SUBTYPES AND SURVIVAL ANALYSIS ANALYSIS OF PRIMARY CENTRAL NERVOUS SYSTEM LYMPHOMA WITH RADIOMICS FEATURES |
|---|---|
| المؤلفون: | BARILLOT, N, HERNANDEZ, I, KIRASIC, E, HOUILLIER, C, MOKHTARI, K, HOANG-XUAN, K, ALENTORN, A |
| المصدر: | Neuro-Oncology ; volume 25, issue Supplement_2, page ii103-ii103 ; ISSN 1522-8517 1523-5866 |
| بيانات النشر: | Oxford University Press (OUP) |
| سنة النشر: | 2023 |
| الوصف: | BACKGROUND Primary Central Nervous System Lymphoma (PCNSL) is a rare and heterogeneous disease with dismal prognosis. Recently, four molecular clusters with clinical relevance have been identified with different potential therapeutic targets in each group. Nevertheless, multi-omics data collection and analysis are expensive and not adapted for clinical practice. Therefore, the identification of surrogate markers to identify PCNSL subtypes from routine data is required, like using hematoxylin and eosin slides from brain biopsies. MATERIAL AND METHODS We used a cohort of 108 patients and we selected the 5000 nuclei for each patient among roughly 1,5M nuclei. Once hematoxylin and eosin slides have been digitized, tessellated, normalized and the nuclei have been segmented and filtered with the computation of a solidity score, the PyRadiomics package provides us with more than 800 features for each nuclei. Firstly, we were interested in survival analysis. In a second time, we also used these features for training classification models. We used a partial least squared Cox model, which is a classic Cox model applied to latent components constructed by using linear combinations of the original variables. RESULTS Results for our first cohort are promising (C-index of 0.87, std 0.01), with a significant increase compared to the clinical features model (C-index of 0.68, std 0.03). We are now challenging these results with three other cohorts of brain and systemic lymphoma. CONCLUSION This study paves the way for a stratification of the clinical evolution based on the machine learning analysis of digital pathology in PCNSL that could be easily translated to a broad range of diseases or other brain tumors. |
| نوع الوثيقة: | article in journal/newspaper |
| اللغة: | English |
| DOI: | 10.1093/neuonc/noad137.346 |
| الاتاحة: | https://doi.org/10.1093/neuonc/noad137.346 https://academic.oup.com/neuro-oncology/article-pdf/25/Supplement_2/ii103/51442456/noad137.346.pdf |
| Rights: | https://academic.oup.com/pages/standard-publication-reuse-rights |
| رقم الانضمام: | edsbas.308F409 |
| قاعدة البيانات: | BASE |
| DOI: | 10.1093/neuonc/noad137.346 |
|---|