Dose Distribution Prediction for Optimal Treamtment of Modern External Beam Radiation Therapy for Nasopharyngeal Carcinoma

Volumetric modulated arc therapy dose prediction and deliverable treatment plan generation for prostate cancer patients using a densely connected deep learning model

“…In this study, we presented a densely connected U-Net, able to predict 3D VMAT dose distributions for prostate cancer patients. In contrast to previous studies, where model training was performed in 2D [1,4,7,11,12], we suggested a 2.5D training approach using image triplets. In addition, we presented a, to the best of our knowledge, novel treatment planning workflow using a commercially available TPS, to transform model predictions into deliverable treatment plans.…”

“…A possible way to improve and automate the optimization and treatment planning process is to use state-of-the-art, machine learningbased algorithms and models. Several research groups are making progress incorporating these new technologies into the EBRT process, to accelerate and improve the treatment planning workflow [1][2][3][4][5][6][7][8][9][10][11][12]. The field of radiation therapy is likely to face tremendous changes regarding patient workflow, availability of artificial intelligence (AI)-based decision supportive tools and treatment planning efficiency and consistency [13][14][15].…”

“…They later refined the model by adding dense connections into a 3D architecture, leading to a more accurate dose prediction [2]. Other studies tried to predict IMRT dose distributions by incorporating dosevolume histogram (DVH) features into the model, or evaluating different architectures, e.g., ResNet-anti-ResNet and generative adversarial networks (GAN) [2][3][4][5][6][7][8].…”

Automatic dose prediction using deep learning and plan optimization with finite‐element control for intensity modulated radiation therapy

“…6 Automatic treatment planning has been investigated intensively, particularly using artificial intelligence to alleviate the clinical burden of medical physicists and improve consistency. [7][8][9] Initially, knowledge-based planning (KBP) emerged for the prediction of dose-volume objectives using different features and algorithms, including organ-to-target distances and support vector regression, 10,11 to expedite the treatment planning. However, the features in KBP, which are compressed from three-dimensional (3D) dosimetric features, yield a loss of spatial information, which cannot ensure the optimal effectiveness of the prediction.…”

“…Automatic treatment planning has been investigated intensively, particularly using artificial intelligence to alleviate the clinical burden of medical physicists and improve consistency 7–9 . Initially, knowledge‐based planning (KBP) emerged for the prediction of dose‐volume objectives using different features and algorithms, including organ‐to‐target distances and support vector regression, 10,11 to expedite the treatment planning.…”