Motion-embedded Deep Learning Model for Local Failure Prediction and Individualized Radiation Dose Prescription (Fellow: Peng T. Troy Teo)

Project: Research project

Project Details


The emergence of conformal, high dose radiation or stereotactic body radiotherapy (SBRT) has significantly altered the management of patients with early-stage lung cancers or oligo-metastatic disease. Despite these advances, there is an emergent category of studies that indicate that certain patients are more likely to experience local failure after SBRT. The poor prognosis is due to the complex interactions between treatment modalities, tumor heterogeneity, and patient-specific variations. Yet, existing radiation treatment planning does not take into account individual patient’s sensitivities or tumor response to radiation. Standard radiation prescription and treatment margins were based on population statistics with the assumption that the tumor is a homogenous mass. To improve outcome, precise radiotherapy (RT) targeting tumors based on their unique genetic make-up and mutations, disease biomarkers, and spatial variations is required. Deep learning radiomics, the process of extracting higher-order image-based features and correlating it with patient data, including biological and clinical endpoints, has been explored for its feasibility in providing precise treatment prognosis. The intratumoral heterogeneity of lung tumors captured with CT radiomics features were found to give prognostic indicators of early-stage lung cancer (P < 0. 01)Preliminary results on the use of radiomics features for inter-fraction RT adaptation, post-RT prognosis for locoregional vs. distant metastases, radiation toxicity prediction, and overall survival prediction are promising. Preliminary results from the work of my proposed supervisor has indicated that deep learning features extracted from CT images are associated with the efficacy of lung SBRT and that prediction models that recommend individualized doses could reduce the probability of local treatment failure to below 5% for lung cancer patients.
Effective start/end date10/1/229/30/24


  • Canadian Institutes of Health Research (AGMT 04/28/22)


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.