Error detection using a convolutional neural network with dose difference maps in patient-specific quality assurance for volumetric modulated arc therapy |
| |
| Institution: | 1. Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan;2. Radiation Oncology Center, Ofuna Chuo Hospital, Kamakura, Japan;3. Department of Radiology, National Hospital Organization Sendai Medical Center, Sendai, Japan;1. Department of Radiology, Takeda General Hospital, Aizuwakamatsu, Japan;2. Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan;3. Department of Radiation Oncology, Tokai University Graduate School of Medicine, Isehara, Japan;4. Department of Radiological Technology, Tohoku University Graduate School of Medicine, Sendai, Japan;1. Department of Radiology, Takeda General Hospital, Aizuwakamatsu, Japan;2. Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan;3. Department of Radiation Oncology, Tokai University School of Medicine, Isehara, Japan;1. Department of Radiology, Takeda General Hospital, Aizuwakamatsu, Japan;2. Department of Radiation Oncology, Tohoku University Graduate School of Medicine, Sendai, Japan;3. Department of Radiation Oncology, Tokai University Graduate School of Medicine, Isehara, Japan;1. Institute of Medical Physics, School of Physics, The University of Sydney, Sydney, New South Wales, Australia;2. School of Medicine, Taif University, Taif, Saudi Arabia;3. Liverpool and Macarthur Cancer Therapy Centers, Liverpool, NSW, Australia;4. Ingham Institute for Applied Medical Research, Sydney, NSW, Australia;5. Centre for Medical Radiation Physics, University of Wollongong, Wollongong, NSW, Australia;6. South Western Sydney Clinical School, University of New South Wales, Sydney, NSW, Australia;1. Department of Radiation Oncology, Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, the Netherlands;2. Department of Radiology and Nuclear Medicine, Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525 GA Nijmegen, the Netherlands |
| |
| Abstract: | The aim of this study was to evaluate the use of dose difference maps with a convolutional neural network (CNN) to detect multi-leaf collimator (MLC) positional errors in patient-specific quality assurance for volumetric modulated radiation therapy (VMAT). A cylindrical three-dimensional detector (Delta4, ScandiDos, Uppsala, Sweden) was used to measure 161 beams from 104 clinical prostate VMAT plans. For the simulation used error-free plans plus plans with two types of MLC error were introduced: systematic error and random error. A total of 483 dose distributions in a virtual cylindrical phantom were calculated with a treatment planning system. Dose difference maps were created from two planar dose distributions from the measured and calculated dose distributions, and these were used as the input for the CNN, with 375 datasets assigned for training and 108 datasets assigned for testing. The CNN model had three convolution layers and was trained with five-fold cross-validation. The CNN model classified the error types of the plans as “error-free,” “systematic error,” or “random error,” with an overall accuracy of 0.944. The sensitivity values for the “error-free,” “systematic error,” and “random error” classifications were 0.889, 1.000, and 0.944, respectively, and the specificity values were 0.986, 0.986, and 0.944, respectively. This approach was superior to those based on gamma analysis. Using dose difference maps with a CNN model may provide an effective solution for detecting MLC errors for patient-specific VMAT quality assurance. |
| |
| Keywords: | Radiotherapy Patient-specific QA Deep learning Volumetric modulated radiation therapy Prostate |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
