Author: Rita Buono, Elisabetta Cagni, Roberta Castriconi, Surendra Bahadur Chand, Marco Esposito, Claudio Fiorino, Valeria Landoni, Aldo Mazzilli, Eugenia Moretti, Lorenzo Placidi, Giulia Rambaldi Guidasci, Alessia Tudda 👨🔬
Affiliation: IRCCS San Raffaele Scientific Institute, Department of Advanced Technology, IRCCS Regina Elena National Cancer Institute, ASU FC Medical Physics, University Hospital of Parma AOUP, ICTP, B.P. Koirala Memorial Cancer Hospital, Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Fatebenefratelli Isola Tiberina – Gemelli Isola 🌍
Purpose: To explore the feasibility and educational impact of transferring knowledge-based planning (KBP) models—developed using Italian breast radiotherapy data—to a Nepalese hospital, thereby demonstrating global knowledge transfer in radiotherapy planning.
Methods: Two benchmark KBP models (for left- and right-sided breast postoperative radiotherapy) were created by the Mikapoco collaboration using data from seven Italian institutions. These RapidPlan (Varian Medical Systems) models employed principal component analysis (PCA) to predict organ-at-risk (OAR) doses for the heart and ipsilateral lung. The models were introduced at B.P. Koirala Memorial Cancer Hospital in Nepal and tested on 20 patients (10 left, 10 right). Transferability was assessed by comparing the first principal component (PC1) of the Nepalese cohort to the 10th–90th percentile range of the Italian dataset. Automated treatment plans were generated using the ViTAT (Virtual Tangential field Arc Therapy) technique for a 40 Gy/15 fractions prescription, then compared to locally produced clinical plans.
Results: Although the higher prevalence of mastectomy in Nepal limited direct model transferability, automated plans remained clinically acceptable and reduced mean OAR doses. For left-sided cases, the ipsilateral lung mean dose decreased from 10.8 ± 2.8 Gy to 10.2 ± 2.4 Gy. Similar reductions were observed for right-sided plans (12 ± 3 Gy vs. 13 ± 3.2 Gy). Target coverage was slightly lower in the KBP plans (e.g., right breast PTV D99% = 58 ± 15% vs. 70 ± 35%), but interpatient variability was reduced, providing a consistent baseline.
Conclusion: Despite differing patient characteristics and treatment volumes, Italian KBP models delivered feasible dose predictions in the Nepalese context. These findings highlight the educational value and global potential of KBP-driven planning, underscoring the need to refine models with diverse patient populations to enhance worldwide applicability and support knowledge transfer.