Abstract 30P
Background
Immune checkpoint inhibitors treatments have proven their efficiency against Non-Small Cell Lung Cancer (NSCLC). However a large number of patients (around 45%) don’t respond to immunotherapies and despite the existing biomarkers (PDL1, CD8) and patient selection practices, there is not currently any relevant predictive biomarker of this response. The recent advances in Deep learning image analysis - namely Multiple Instance Learning - allow to train a model to predict survival for patients receiving immunotherapy and to confront visual results to known biomarkers.
Methods
2220 H&E histology images were gathered from 456 NSCLC patients. The 36-month follow-up with vital status is retrospectively documented (68.5% alive, and 31.5% deceased). The slides are first tiled into patches and histology phenotypes are extracted using K-means on VGG-16 feature extractor. The phenotypes are then fed into siamese networks and aggregated into a single patient representation through an Attention layer predicting the vital status (alive / deceased). The Attention map is finally used to extract the most important phenotypes driving the model decision of immunotherapy response. The result of the attention maps is compared with the CD8 and PDL1 biomarker stainings.
Results
We find that a relevant number of histologically phenotypes used for the analysis is 4 and that, based on tiles coming from each of these four phenotypes the survival prediction can be performed with an AUC of 0.75. Qualitative analysis shows that the model decision for predicting response to immunotherapy from H&E stained slide not only rely on PDL1 biomarker but also how immune cells infiltrate the tumoral PDL1 area.
Conclusions
In this study we use a new deep learning approach enabling us to discover new histological phenotypes that predict the response to immunotherapy in NSCLC. Our first results are encouraging and demonstrate the AI capability for discovering new discriminative phenotypes. Our future work will be focused on inspecting more granular phenotypes.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Tribun Health.
Disclosure
All authors have declared no conflicts of interest.
Resources from the same session
86P - Validation of the Gustave Roussy Immune (GRIm) score in patients treated with bispecific CD3 T cell engagers in phase I clinical trials
Presenter: Noé Herbel
Session: Cocktail & Poster Display session
Resources:
Abstract
87P - Clinical features and genetic profile of MDM2-altered solid tumors
Presenter: Julia Tejerina-Peces
Session: Cocktail & Poster Display session
Resources:
Abstract
88P - Evaluation of trending drug targets and technologies in current drug development
Presenter: Matteo Repetto
Session: Cocktail & Poster Display session
Resources:
Abstract
89P - Impact of metformin on glucocorticoid-induced changes in systemic metabolism in patients with brain metastases from solid malignancies
Presenter: Filippo Guglielmo Maria De Braud
Session: Cocktail & Poster Display session
Resources:
Abstract
90P - Effect of age on safety and efficacy of novel cancer drugs investigated in early phase clinical trials
Presenter: Jacopo Uliano
Session: Cocktail & Poster Display session
Resources:
Abstract
91P - Molecular predictors of response to the therapy with mitotane in adrenocortical cancer
Presenter: Erika Porubayeva
Session: Cocktail & Poster Display session
Resources:
Abstract
93P - Circulating cell-free DNA fragmentation profiles during systemic therapy of advanced-stage non-small cell lung cancer patients
Presenter: Jelena Milovanovic
Session: Cocktail & Poster Display session
Resources:
Abstract
94P - Targeted therapy in pediatric acute myeloid leukemia: An unmet need
Presenter: Banda Teja
Session: Cocktail & Poster Display session
Resources:
Abstract
95P - Anticancer activity of Inula recemosa root extract in human liver cancer cell line by attenuation of OCT4/Sox2 axis
Presenter: Tania Ghosh(Sarkar)
Session: Cocktail & Poster Display session
Resources:
Abstract
96P - The Hammersmith score optimises patient selection and predicts for overall survival in early-phase cancer trial participants independent of tumour burden
Presenter: James Korolewicz
Session: Cocktail & Poster Display session
Resources:
Abstract