A paper on automatic safety verification technology (virtual patient QA) for IMRT using AI has been published online!
This paper shows the usefulness of a new method for automatic safety verification of IMRT using AI.Title: Improvement of deep learning prediction model in patient-specific QA for VMAT with MLC leaf position map and patient’s dose distribution
Ryota Tozuka, Noriyuki Kadoya, Seiji Tomori, Yuto Kimura, Tomohiro Kajikawa, Yuto Sugai, Yushan Xiao, Keiichi Jinguhttps://doi.org/10.1002/acm2.14055Technical explanation
A technology that predicts the measured values for safety verification with high accuracy by inputting the “dose distribution in the patient” and the “MLC motion map” of the radiotherapy planning data into deep learning.
By using this technology, it is possible to virtually verify the safety of IMRT using AI without the conventional safety verification using a detector (patient-specific QA). This makes it possible to improve the efficiency of safety verification (about 1 hour → 1 minute).
IMRT treatment planning can create a steep dose distribution because multiple metal plates that shield radiation called multi-leaf collimator (MLC) can be moved in a complicated manner to irradiate. On the other hand, it is necessary to verify in advance whether irradiation can be performed safely as planned before irradiation to the patient. Complex MLC movement creates an extreme dose distribution, so if the MLC does not move as planned, an unexpected dose may be delivered to dangerous organs. However, one issue is that this verification requires one hour of verification time per case, which is a major factor in increasing the workload of medical staff.Furthermore, in recent years, a technique called online adaptive radiotherapy (Online-ART), in which the treatment plan is modified according to the patient’s condition for each treatment session, has begun to spread. With this Online-ART, a treatment plan is created immediately while the patient is on the treatment table, and irradiation is performed according to that treatment plan. Therefore, it is not possible to carry out actual measurement verification using conventional detectors. Therefore, streamlining the validation of this treatment plan has become an urgent issue.
An example of a device that can perform Online-ART is a device called MR-Linac, which integrates an MR device and a linac