Jan C Peeken1,2, Tatyana Goldberg3, Christoph Knie4, Basil Komboz3, Michael Bernhofer5, Francesco Pasa6,7, Kerstin A Kessel4,8,9, Pouya D Tafti3, Burkhard Rost5, Fridtjof Nüsslin4, Andreas E Braun3, Stephanie E Combs4,8,9. 1. Department of Radiation Oncology, Klinikum rechts der Isar, Technical University of Munich (TUM), Ismaninger Straße 22, 81675, Munich, Germany. jan.peeken@tum.de. 2. Partner Site Munich, Deutsches Konsortium für Translationale Krebsforschung (DKTK), Munich, Germany. jan.peeken@tum.de. 3. Allianz SE, Königinstraße 28, 80802, Munich, Germany. 4. Department of Radiation Oncology, Klinikum rechts der Isar, Technical University of Munich (TUM), Ismaninger Straße 22, 81675, Munich, Germany. 5. Department for Bioinformatics and Computational Biology, Informatik 12, Technical University of Munich (TUM), Boltzmannstraße 3, 85748, Garching, Germany. 6. Department of Computer Science, Informatik 9, Technical University of Munich (TUM), Boltzmannstraße 3, 85748, Garching, Germany. 7. Chair of Biomedical Physics, Department of Physics, Technical University of Munich (TUM), James-Franck-Straße 1, 85748, Garching, Germany. 8. Institute of Innovative Radiotherapy (iRT), Department of Radiation Sciences (DRS), Helmholtz Zentrum München, Ingolstaedter Landstraße 1, 85764, Neuherberg, Germany. 9. Partner Site Munich, Deutsches Konsortium für Translationale Krebsforschung (DKTK), Munich, Germany.
Abstract
BACKGROUND AND PURPOSE: Current prognostic models for soft tissue sarcoma (STS) patients are solely based on staging information. Treatment-related data have not been included to date. Including such information, however, could help to improve these models. MATERIALS AND METHODS: A single-center retrospective cohort of 136 STS patients treated with radiotherapy (RT) was analyzed for patients' characteristics, staging information, and treatment-related data. Therapeutic imaging studies and pathology reports of neoadjuvantly treated patients were analyzed for signs of response. Random forest machine learning-based models were used to predict patients' death and disease progression at 2 years. Pre-treatment and treatment models were compared. RESULTS: The prognostic models achieved high performances. Using treatment features improved the overall performance for all three classification types: prediction of death, and of local and systemic progression (area under the receiver operatoring characteristic curve (AUC) of 0.87, 0.88, and 0.84, respectively). Overall, RT-related features, such as the planning target volume and total dose, had preeminent importance for prognostic performance. Therapy response features were selected for prediction of disease progression. CONCLUSIONS: A machine learning-based prognostic model combining known prognostic factors with treatment- and response-related information showed high accuracy for individualized risk assessment. This model could be used for adjustments of follow-up procedures.
BACKGROUND AND PURPOSE: Current prognostic models for soft tissue sarcoma (STS) patients are solely based on staging information. Treatment-related data have not been included to date. Including such information, however, could help to improve these models. MATERIALS AND METHODS: A single-center retrospective cohort of 136 STSpatients treated with radiotherapy (RT) was analyzed for patients' characteristics, staging information, and treatment-related data. Therapeutic imaging studies and pathology reports of neoadjuvantly treated patients were analyzed for signs of response. Random forest machine learning-based models were used to predict patients' death and disease progression at 2 years. Pre-treatment and treatment models were compared. RESULTS: The prognostic models achieved high performances. Using treatment features improved the overall performance for all three classification types: prediction of death, and of local and systemic progression (area under the receiver operatoring characteristic curve (AUC) of 0.87, 0.88, and 0.84, respectively). Overall, RT-related features, such as the planning target volume and total dose, had preeminent importance for prognostic performance. Therapy response features were selected for prediction of disease progression. CONCLUSIONS: A machine learning-based prognostic model combining known prognostic factors with treatment- and response-related information showed high accuracy for individualized risk assessment. This model could be used for adjustments of follow-up procedures.
Entities:
Keywords:
Biomarker; Decision support systems; Precision medicine; Prognostic model; Random forest
Authors: H D Suit; H J Mankin; W C Wood; M C Gebhardt; D C Harmon; A Rosenberg; J E Tepper; D Rosenthal Journal: J Clin Oncol Date: 1988-05 Impact factor: 44.544
Authors: Robert G Maki; Nicole Moraco; Cristina R Antonescu; Meera Hameed; Alisa Pinkhasik; Samuel Singer; Murray F Brennan Journal: Ann Surg Oncol Date: 2013-06-18 Impact factor: 5.344
Authors: Jens Jakob; Tom Lesluyes; Anna Simeonova-Chergou; Frederik Wenz; Peter Hohenberger; Frederic Chibon; Sophie Le Guellec Journal: Strahlenther Onkol Date: 2019-11-15 Impact factor: 3.621
Authors: Jan C Peeken; Tatyana Goldberg; Thomas Pyka; Michael Bernhofer; Benedikt Wiestler; Kerstin A Kessel; Pouya D Tafti; Fridtjof Nüsslin; Andreas E Braun; Claus Zimmer; Burkhard Rost; Stephanie E Combs Journal: Cancer Med Date: 2018-12-18 Impact factor: 4.452
Authors: Jan C Peeken; Matthew B Spraker; Carolin Knebel; Hendrik Dapper; Daniela Pfeiffer; Michal Devecka; Ahmed Thamer; Mohamed A Shouman; Armin Ott; Rüdiger von Eisenhart-Rothe; Fridtjof Nüsslin; Nina A Mayr; Matthew J Nyflot; Stephanie E Combs Journal: EBioMedicine Date: 2019-09-12 Impact factor: 8.143