Journal of applied clinical medical physics · 2025
Evaluation of deep learning-based automated radiotherapy planning for early-stage lung cancer using SBRT-VMAT: A comparison with manual planning.
СтатьяСравнительное исследованиеОценочное исследованиеНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Nemoto Hikaru, Saito Masahide, Kadoya Noriyuki, Komiyama Takafumi, Tozuka Ryota, Onishi Hiroshi, Jingu Keiichi
Аннотация
PURPOSE: This study evaluates the feasibility of deep learning (DL)-based automated SBRT-VMAT planning for lung cancer.
METHODS: We developed a DL-based dose distribution prediction model (installed in the RatoGuide prototype, AiRato. Inc.), which was trained using manually optimized and clinically approved plans for 124 cases of early-stage lung cancer (62 central and 62 peripheral cases). The test data consisted of eight treatment plans for each case. Automated plans for the test data were created as follows: (i) predicting dose distributions from CT images with target and organs-at-risk contours using RatoGuide, (ii) converting the predicted dose distributions into ring-shaped dose structures and importing them into RayStation software, and (iii) generating deliverable dose distributions automatically to reproduce the predicted dose distributions using a scripting application within the treatment planning system. We measured the time required to create the plans and compared the DVH metrics of the automated and manual plans. Two expert radiation oncologists evaluated the automated and manual plans based on following aspects: (a) clinical acceptability and (b) preference for automated or manual plans in clinical practice.
RESULTS: No significant differences were observed in any DVH metric between the automated and manual plans in each case. Two radiation oncologists reviewed all automated plans and deemed them clinically acceptable. The number of cases evaluated as the preferred automated plan by radiation oncologists was higher for central than peripheral cases.
CONCLUSIONS: These results indicate the feasibility of automated treatment planning using a prediction model based on DL, suggesting that DL techniques can efficiently generate clinically acceptable SBRT-VMAT treatment plans for lung cancer.
METHODS: We developed a DL-based dose distribution prediction model (installed in the RatoGuide prototype, AiRato. Inc.), which was trained using manually optimized and clinically approved plans for 124 cases of early-stage lung cancer (62 central and 62 peripheral cases). The test data consisted of eight treatment plans for each case. Automated plans for the test data were created as follows: (i) predicting dose distributions from CT images with target and organs-at-risk contours using RatoGuide, (ii) converting the predicted dose distributions into ring-shaped dose structures and importing them into RayStation software, and (iii) generating deliverable dose distributions automatically to reproduce the predicted dose distributions using a scripting application within the treatment planning system. We measured the time required to create the plans and compared the DVH metrics of the automated and manual plans. Two expert radiation oncologists evaluated the automated and manual plans based on following aspects: (a) clinical acceptability and (b) preference for automated or manual plans in clinical practice.
RESULTS: No significant differences were observed in any DVH metric between the automated and manual plans in each case. Two radiation oncologists reviewed all automated plans and deemed them clinically acceptable. The number of cases evaluated as the preferred automated plan by radiation oncologists was higher for central than peripheral cases.
CONCLUSIONS: These results indicate the feasibility of automated treatment planning using a prediction model based on DL, suggesting that DL techniques can efficiently generate clinically acceptable SBRT-VMAT treatment plans for lung cancer.