Upcoming Planned Work:
- Multimodal clinical and radiomic model validation on external dataset
- Feasibility assessment of multimodal clinical, radiomic and genomic machine learning model using a similarity network fusion approach
- Dissemination of results and lessons learne
Multi-omic Assessment of Squamous cell cancers receiving Systemic Therapy (MASST)
Co-PIs: Elena Elimova, Kirsty Taylor, Lillian Siu
Multimodal models bring together distinct datasets for a patient that span modalities and provide the opportunity for a more comprehensive understanding of the whole patient that is not possible with a single type of data. Multimodal data, which may include data types such as clinical, radiomic and genomic, can be used in machine learning models that provide more information than standard one-dimensional models. By combining these traditionally disparate datasets into a single model, new patient-specific insights are possible.
In the context of the patient population considered, using these insights to proactively identify patients with recurrent metastatic Head and Neck squamous cell cancers who would most benefit from immunotherapy, could lead to better outcomes for patients.
Clinical, radiomic and genomic data have been collected for study patients, and clinical and radiomic multimodal machine learning (ML) models have been developed and tested. Genomic modeling is underway using circulating tumor DNA (ctDNA) data.
- All clinical and radiomic data collected for the MASST study including approximately 4000 clinical data points, 1200 radiomic features. Genomic data collection underway.
- Clinical, radiomic and genomic workflows mapped from data collection to final analysis
- Radiomic machine learning model externally validated on MASST dataset.
- Clinical machine learning model developed and tested
- Multimodal analysis framework/methodology selected
- Validation completed for manually collected study data versus data pulls from UHN systems via UHN’s Digital Health Platform (DHP)
Last modified: May 7, 2021