Our latest work

Happy to announce that our work Deep Evidential Learning for Radiotherapy Dose Prediction has been accepted for publication in the Tier-1 Computer Science journal Computers in Biology and Medicine !

• read our openly accessible preprint at https://arxiv.org/abs/2404.17126 

Here’s the graphical abstract prepared for the journal publication: 

In this work, we presented a deep learning model to predict the radiation dose given the patient’s set of CT images showing various organs and target areas delineated by the radiation oncologist. It also encapsulates an uncertainty-quantification framework where we can put confidence intervals on Dose-Volume-Histograms and use uncertainty heatmaps to discover potential areas of model errors. 

Our work extends previous work on using deep learning models to predict the radiotherapy dose distribution given the targeted region of the patient’s anatomy and CT images. This can be further used as input for an A.I.-automated approach towards designing the set of radiotherapy beam and related machine parameters for the treatment of the patient. From the computer science perspective, we performed a refinement of an uncertainty quantification framework known as Deep Evidential Learning so that it can be incorporated into a standard convolutional neural network to implement the radiation dose prediction. For more details, check out our preprint and a more detailed description of the project under our website section on Research Projects.