I.D. Jonkhof
Research Subjects
I.D. Jonkhof has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Erik. W. Korevaar
Department of Radiotherapy, Universitair Medisch Centrum Groningen, PO Box 30 001, Groningen, 9700 RB, NETHERLANDS
1 shared publication - Ilse van Bruggen1 shared publication
- Johannes Albertus Langendijk
Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands
1 shared publication Minke Brinkman-Akker
1 shared publication- Stefan Both
University Medical Center Groningen
1 shared publication
Latest Publications
- Open Access
- 12 March 2026
- Published by Elsevier in SSRN Electronic Journal
Monitoring plan quality in clinically integrated deep learning-based IMPT planning for oropharyngeal cancer patients
Background and Purpose: With deep learning becoming an integral component of radiotherapy planning, it is critical to verify its consistency during clinical deployment. This study monitored the plan quality of clinically integrated deep learning (DL)–based planning for generating robustly optimized intensity modulated proton therapy (IMPT) plans in oropharyngeal carcinoma. Materials and methods: A U net–based DL model in the RayStation planning system predicted dose distributions, followed by robust dose mimicking to generate a standard DL optimized plan (sDLO). Dosimetrists refined the sDLO to create the manually adjusted DLO (mDLO) plan used for treatment. Plan quality of all mDLO plans was evaluated for 78 patients in a multidisciplinary meeting. To assess long term performance across the 28 month routine clinical use, manual benchmark plans were generated for 16 patients (every fifth patient). Quantitative evaluation included target coverage (CTV D98 ≥ 94 %, voxel wise minimum and maximum), organ at risk (OAR) doses, and normal tissue complication probabilities (NTCP) for grade ≥2 dysphagia and xerostomia. Results: All 78 mDLO plans (100%) were clinically acceptable. In the benchmark cohort, 14/16 mDLO plans (87.5%) met all clinical goals. Mean parotid dose was significantly lower in mDLO plans compared with manual plans (15.8 Gy (RBE) vs 16.9 Gy (RBE), p = 0.002). Manual refinement increased the number of patients meeting all clinical goals from 59 to 68. Conclusion: Clinically integrated deep learning–based IMPT planning consistently produced high‑quality, clinically acceptable plans for all 78 patients, with plan quality maintained during 28 months of routine clinical use.
- 12 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMonitoring plan quality in clinically integrated deep learning-based IMPT planning for oropharyngeal cancer patients
Background and Purpose: With deep learning becoming an integral component of radiotherapy planning, it is critical to verify its consistency during clinical deployment. This study monitored the plan quality of clinically integrated deep learning (DL)–based planning for generating robustly optimized intensity modulated proton therapy (IMPT) plans in oropharyngeal carcinoma. Materials and methods: A U net–based DL model in the RayStation planning system predicted dose distributions, followed by robust dose mimicking to generate a standard DL optimized plan (sDLO). Dosimetrists refined the sDLO to create the manually adjusted DLO (mDLO) plan used for treatment. Plan quality of all mDLO plans was evaluated for 78 patients in a multidisciplinary meeting. To assess long term performance across the 28 month routine clinical use, manual benchmark plans were generated for 16 patients (every fifth patient). Quantitative evaluation included target coverage (CTV D98 ≥ 94 %, voxel wise minimum and maximum), organ at risk (OAR) doses, and normal tissue complication probabilities (NTCP) for grade ≥2 dysphagia and xerostomia. Results: All 78 mDLO plans (100%) were clinically acceptable. In the benchmark cohort, 14/16 mDLO plans (87.5%) met all clinical goals. Mean parotid dose was significantly lower in mDLO plans compared with manual plans (15.8 Gy (RBE) vs 16.9 Gy (RBE), p = 0.002). Manual refinement increased the number of patients meeting all clinical goals from 59 to 68. Conclusion: Clinically integrated deep learning–based IMPT planning consistently produced high‑quality, clinically acceptable plans for all 78 patients, with plan quality maintained during 28 months of routine clinical use.