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
24P - Fibroblast growth factor receptor (FGFR) as a target for epigenetic therapies: A case review for a new approach to target tumors with somatic mutations/amplifications of FGFR
Presenter: M. Nezami
Session: Cocktail & Poster Display session
Resources:
Abstract
25P - Relationship between aberrant methylation of CpG islands of microRNA gene promoters and changes in their expression in epithelial ovarian cancer
Presenter: Irina Pronina
Session: Cocktail & Poster Display session
Resources:
Abstract
26P - 3MST: A potential workhorse in H2S signaling trimmed by microRNA-548 in breast cancer
Presenter: Alyaa Dawoud
Session: Cocktail & Poster Display session
Resources:
Abstract
29P - A phase I, open-label, dose escalation and dose expansion study to evaluate the safety, tolerability, pharmacokinetics/pharmacodynamics, antitumor activity of QL1604, a humanized anti-PD-1 mAb, in patients with advanced solid tumors
Presenter: Yun Fan
Session: Cocktail & Poster Display session
Resources:
Abstract
32P - Clinical efficacy and safety of an immune checkpoint inhibitor in combination with regorafenib therapy as second-line regimen for patients with unresectable hepatocellular carcinoma
Presenter: Jinpeng Li
Session: Cocktail & Poster Display session
Resources:
Abstract
33P - Nomogram to predict survival of patients with unresectable melanoma receiving immune checkpoint inhibitors
Presenter: Eftychia Chatziioannou
Session: Cocktail & Poster Display session
Resources:
Abstract
34P - Targeting LSD1 rescues MHC-I antigen presentation and overcomes resistance to PD-L1 checkpoint blockade in small cell lung cancer
Presenter: Evelyn Nguyen
Session: Cocktail & Poster Display session
Resources:
Abstract
36P - Therapeutic vaccination with HPV-16 oncoproteins fused into a checkpoint modifier of early T cell activation protects against HPV-associated tumors in a preclinical model
Presenter: Susan Currie
Session: Cocktail & Poster Display session
Resources:
Abstract
37P - Dose transition pathways for time-to-event continual reassessment method (TITE-CRM): Will imposing a waiting time result in better performance?
Presenter: Zhulin Yin
Session: Cocktail & Poster Display session
Resources:
Abstract
38P - Locoregional radiotherapy improves survival outcomes in de novo metastatic nasopharyngeal carcinoma treated with chemoimmunotherapy
Presenter: Yujun Hu
Session: Cocktail & Poster Display session
Resources:
Abstract