A personalized DVH prediction model for HDR brachytherapy in cervical cancer treatment

• DOI: 10.3389/fonc.2022.967436
• Publication Date: 2022-08-30T08:51:06Z
• Authors: Zhen Li; Kehui Chen; Zhenyu Yang; Qingyuan Zhu; Xiaojing Yang; Zhaobin Li; Jie Fu
• Keywords: Dose-volume histogram

Although the knowledge-based dose-volume histogram (DVH) prediction has been largely researched and applied in External Beam Radiation Therapy, it is still less investigated domain of brachytherapy. The purpose this study to develop a reliable DVH method for high-dose-rate brachytherapy plans.A workflow combining kernel density estimation (KDE), k-nearest neighbor (kNN), principal component analysis (PCA) was proposed. PCA kNN were first employed together select similar patients based on directions. 79 cervical cancer with different applicators inserted included study. KDE model built relationship between distance-to-target (DTH) dose selected cases, which can be subsequently used estimate probability distribution validation set. Model performance bladder rectum quantified by |ΔD2cc|, |ΔD1cc|, |ΔD0.1cc|, |ΔDmax|, |ΔDmean| form mean standard deviation. only combination kNN, PCA, compared.20, 30 KNN respectively. absolute residual actual plans predicted 0.38 ± 0.29, 0.4 0.32, 0.43 0.36, 0.97 0.66, 0.13 0.99 bladder, For rectum, corresponding results 0.34 0.27, 0.33, 0.63 0.57, 1.41 0.23 0.17, showed significantly better than only, an improvement 30.3% 33.3% rectum.In study, machine learning proposed verified accurately predict new patients. This proved effective our testing group HDR