​Intensity modulated radiotherapy, or IMRT, represents a significant advance over conventional 3D conformal radiotherapy. With IMRT, the dose prescribed can be delivered in a targeted manner to the tumor, limiting exposure to adjacent organs and, therefore, lowering potential side effects of treatment. But before a patient can undergo IMRT, strict quality controls must be completed to make sure that the treatment plan will be safe and accurate.

The purpose of the treatment plan is to determine how to deliver the right dose of radiation to the right location by configuring the beam trajectory, shape, speed, angle, and shooting time. Not having to check all parameters for every treatment plan would save medical physicists precious time. New decision-support software developed by CEA-List can predict the likelihood that a treatment plan will pass these quality controls. This initial analysis can be used to automatically eliminate or approve certain treatment plans and to make controlling the quality of remaining plans more efficient.

The researchers who came up with the tool leveraged Bayesian techniques to develop novel classification algorithms and statistical learning models. The software is trained on databases containing the characteristics of treatment plans previously used on patients so that a complexity rating can be assigned to new plans and the likelihood that they would pass quality control can be predicted. And, because the software also provides an indication of how accurate the prediction is, it supports better-informed decision-making by medical physicists. CE marking for the tool is pending, and it has already been licensed to radiotherapy clinic operator RT3C for deployment at its facilities, confirming CEA-List's role as an active contributor to the field of digital health.