Purpose: We proposed an artificial neural network (ANN) model to predict radiobiological parameters for head and neck squamous cell carcinoma (HNSCC) patients who are receiving radiation therapy. The model predicts the tumor control probability (TCP) and normal tissue complication probability (NTCP). These parameters play an important role during treatment planning evaluation and clinical management of diseases.
Methods: A publicly available dataset of 31 HNSCC patients underwent highly conformal radiation therapy was selected. The dataset includes contours of organ at risk and planning target volumes for radiation therapy, outcome/follow-up measurements and dose-volume histograms for each delivered plan. We extracted patients’ demographics, cancer stage and performance status, and dose-volume histograms from the dataset. These parameters were used as inputs for our ANN model. RADBIOMOD v.3.0 (a spreadsheet-based freeware for calculation of radiobiological indices) was used to calculate TCP and NTCP for all the delivered plans and their values were used as output. The distribution of data for training, validation and testing purpose of ANN are 70%, 15% and 15% respectively.
Results: We found the regression value for the prediction of TCP in stages of training, validation, test and overall are 0.80, 0.86, 0.95 and 0.82 respectively. Similarly, for NTCP the values are 0.93, 0.89, 0.86 and 0.88.
Conclusion: We believe that the model has significant potential to predict radiobiological indices and help clinicians in treatment plan evaluation and treatment management of the HNSCC patients.
TH- Radiobiology(RBio)/Biology(Bio): Rbio - Outcome models combining dose, imaging, radiomics/radiogenomics and clinical factors: machine learning