Zusammenfassung
Hintergrund
Die Bestrahlungsplanung ist ein notwendiger Arbeitsschritt vor Durchführung einer Strahlentherapie. Die Definition von anatomischen Organen, die in direkter Nähe zu der bestrahlten Zielregion liegen, und die Definition des Zielvolumens ist dabei ein zentraler Bestandteil der ärztlichen Tätigkeit einer Strahlentherapeut*In. Die Errungenschaften in der Entwicklung der künstlichen Intelligenz (KI) haben neuronale Netze hervorgebracht, die hocheffektiv zur Segmentierung von medizinischen Bilddaten verwendet werden können.
Ziel der Arbeit
Ziel war die Analyse der Möglichkeiten der KI-basierten Autokonturierung in der Bestrahlungsplanung. Dabei erfolgt die Vorstellung von wissenschaftlichen Arbeiten, die Diskussion klinisch verfügbarer Software und ein Ausblick auf zukünftige innovative Lösungen.
Material und Methoden
Eine Literatursuche (PubMed) wurde durchgeführt, um relevante Literatur zu identifizieren.
Ergebnisse
Erste zugelassene Softwarelösungen ermöglichen die automatisierte Konturierung von anatomischen Organen. Die Segmentierungsgüte erreicht für viele Organe eine hohe Qualität, während bestimmte kleine oder besonders lagevariable Strukturen noch größerer manueller Korrekturen bedürfen. Die Definition von klinischen Zielvolumina, z. B. im Sinne von lokalen Lymphabflusswegen, scheint eine gute Reproduzierbarkeit aufzuweisen. Für verschiedene Tumoren wurde außerdem gezeigt, dass neuronale Netze ebenfalls effektiv und reproduzierbar die Tumorregion definieren können. Weitere Entwicklungen, wie Tumorwachstumsmodelle, könnten außerdem neue individualisierte Definitionswege von Zielvolumina ermöglichen.
Schlussfolgerung
KI-Modelle zur Autokonturierung haben das Potenzial, die Arbeit von Radioonkolog*Innen durch eine Teilautomatisierung zu beschleunigen, den Personalaufwand zu reduzieren und gleichzeitig eine erhöhte Standardisierung zu erreichen.
Abstract
Background
Target volume definition is a central component of radiation planning in radiation oncology. In addition to anatomical organs that are in close proximity to the irradiated target region, target volume definition is a relevant part of a radiation oncologist’s medical practice. Advances in the development of artificial intelligence (AI) have produced neural networks that can be used highly effectively to segment medical image data.
Aim
To analyze the potential of AI-based autocontouring in radiation planning. The article presents the body of scientific work, existing software solutions and an outlook on future innovative solutions.
Materials and methods
A literature search (PubMed) was performed to identify relevant literature.
Results
The first approved software solutions allow automated contouring of anatomical organs. The segmentation quality for many organs is high, while certain positionally variable structures or particularly small organs still require more substantial corrections. The definition of clinical target volumes, e.g., in terms of local lymphatic drainage, achieve good reproducibility. For a variety of tumors, it has also been shown that neural networks can effectively and reproducibly define the gross tumor volume. Further developments, such as tumor growth models, may provide novel individualized definitions of target volumes.
Conclusion
AI models for autocontouring have the potential to accelerate the work of radiation oncologists through partial automation and reducing staff time, while achieving increased standardization.
Literatur
Wannenmacher M, Debus J, Wenz F (Hrsg) (2006) Strahlentherapie. Springer, Berlin, Heidelberg
LeCun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521:436–444
Truhn D, Schrading S, Haarburger C, Schneider H, Merhof D, Kuhl C (2019) Radiomic versus Convolutional Neural Networks Analysis for Classification of Contrast-enhancing Lesions at Multiparametric Breast MRI. Radiology 290:290–297
Peeken JC, Wiestler B, Combs SE (2020) Image-Guided Radiooncology: The Potential of Radiomics in Clinical Application. In Debus J, Schober O, Kiessling F (eds) Image Guided Radiooncology. Springer Berlin Heidelberg, p. 773–94.
Ronneberger O, Philipp F, U‑Net BT (2015) Convolutional Networks for Biomedical Image Segmentation. MICCAI 9351:234–241
Milletari F, Navab N, Ahmadi S‑A. V‑Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation. In: 2016 Fourth International Conference on 3D Vision (3DV). IEEE; 2016.
Isensee F, Jaeger PF, Kohl SAA, Petersen J, Maier-Hein KH (2021) nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods 18:203–211
Navarro F, Shit S, Ezhov I et al (2019) Shape-Aware Complementary-Task Learning for Multi-organ Segmentation. In: MLMI Workshop 2019 MICCAI, S 620–627
Navarro F, Sasahara G, Shit S, et al. A Unified 3D Framework for Organs-at-Risk Localization and Segmentation for Radiation Therapy Planning. In: 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). IEEE; 2022.
Zabel WJ, Conway JL, Gladwish A et al (2021) Clinical Evaluation of Deep Learning and Atlas-Based Auto-Contouring of Bladder and Rectum for Prostate Radiation Therapy. Pract Radiat Oncol 11:e80–9
Strolin S, Santoro M, Paolani G et al (2023) How smart is artificial intelligence in organs delineation? Testing a CE and FDA-approved Deep-Learning tool using multiple expert contours delineated on planning CT images. Front Oncol 13:1–13
Thor M, Apte A, Haq R, Iyer A, LoCastro E, Deasy JO (2021) Using Auto-Segmentation to Reduce Contouring and Dose Inconsistency in Clinical Trials: The Simulated Impact on RTOG 0617. Int J Radiat Oncol Biol Phys 109:1619–1626
Buchner JA, Kofler F, Etzel L et al (2023) Development and external validation of an MRI-based neural network for brain metastasis segmentation in the AURORA multicenter study. Radiother Oncol 178:109425
Primakov SP, Ibrahim A, van Timmeren JE et al (2022) Automated detection and segmentation of non-small cell lung cancer computed tomography images. Nat Commun 13:3423
Song Y, Hu J, Wu Q et al (2020) Automatic delineation of the clinical target volume and organs at risk by deep learning for rectal cancer postoperative radiotherapy. Radiother Oncol 145:186–192
Wong J, Fong A, McVicar N et al (2020) Comparing deep learning-based auto-segmentation of organs at risk and clinical target volumes to expert inter-observer variability in radiotherapy planning. Radiother Oncol 144:152–158
Schreier J, Genghi A, Laaksonen H, Morgas T, Haas B (2020) Clinical evaluation of a full-image deep segmentation algorithm for the male pelvis on cone-beam CT and CT. Radiother Oncol 145:1–6
Kucharczyk M, Chytyk-Praznik K, Giambattista J et al (2022) 18: A Randomized Blinded Assessment of a Machine Learning Based Autocontouring Tool for Breast Cancer Radiotherapy Compared to Peer-Reviewed Radiation Oncologist Contours. Radiother Oncol 174:S11
Shusharina N, Söderberg J, Edmunds D, Löfman F, Shih H, Bortfeld T (2020) Automated delineation of the clinical target volume using anatomically constrained 3D expansion of the gross tumor volume. Radiother Oncol 146:37–43
Peeken JC, Shouman MA, Kroenke M et al (2020) A CT-based radiomics model to detect prostate cancer lymph node metastases in PSMA radioguided surgery patients. Eur J Nucl Med Mol Imaging 47:2968–2977
Rathore S, Akbari H, Doshi J (2018) Radiomic signature of infiltration in peritumoral edema predicts subsequent recurrence in glioblastoma: implications for personalized radiotherapy planning. J Med Imaging 5:1
Peeken JC, Molina-Romero M, Diehl C et al (2019) Deep learning derived tumor infiltration maps for personalized target definition in Glioblastoma radiotherapy. Radiother Oncol 138:166–172
Lipkova J, Angelikopoulos P, Wu S et al (2019) Personalized Radiotherapy Design for Glioblastoma: Integrating Mathematical Tumor Models, Multimodal Scans, and Bayesian Inference. IEEE Trans Med Imaging 38:1875–1884
Metz M‑C, Ezhov I, Zimmer L et al (2023) Towards Image-Based Personalization of Glioblastoma Therapy A Clinical and Biological Validation Study of a Novel, Deep Learning-Driven Tumor Growth Model, S 1–19
Spohn SKB, Schmidt-Hegemann N‑S, Ruf J et al (2023) Development of PSMA-PET-guided CT-based radiomic signature to predict biochemical recurrence after salvage radiotherapy. Eur J Nucl Med Mol Imaging. https://doi.org/10.1007/s00259-023-06195-3
Spohn SKB, Farolfi A, Schandeler S et al (2022) The maximum standardized uptake value in patients with recurrent or persistent prostate cancer after radical prostatectomy and PSMA-PET-guided salvage radiotherapy—a multicenter retrospective analysis. Eur J Nucl Med Mol Imaging 50:218–227
Peeken JC, Asadpour R, Specht K et al (2021) MRI-based delta-radiomics predicts pathologic complete response in high-grade soft-tissue sarcoma patients treated with neoadjuvant therapy. Radiother Oncol 164:73–82
Peeken JC, Neumann J, Asadpour R et al (2021) Prognostic Assessment in High-Grade Soft-Tissue Sarcoma Patients: A Comparison of Semantic Image Analysis and Radiomics. Cancers (Basel) 13:1929
Lang DM, Peeken JC, Combs SE, Wilkens JJ, Bartzsch S (2021) Deep Learning Based HPV Status Prediction for Oropharyngeal Cancer Patients. Cancers (Basel) 13:786
Navarro F, Dapper H, Asadpour R et al (2021) Development and External Validation of Deep-Learning-Based Tumor Grading Models in Soft-Tissue Sarcoma Patients Using MR Imaging. Cancers (basel) 13:2866
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Interessenkonflikt
J. Peeken und S. Combs geben an, dass kein Interessenkonflikt besteht.
Für diesen Beitrag wurden von den Autor/-innen keine Studien an Menschen oder Tieren durchgeführt. Für die aufgeführten Studien gelten die jeweils dort angegebenen ethischen Richtlinien.
Additional information
QR-Code scannen & Beitrag online lesen
Rights and permissions
About this article
Cite this article
Peeken, J.C., Combs, S.E. Anwendung künstlicher Intelligenz in der Radioonkologie. Onkologie 29, 876–882 (2023). https://doi.org/10.1007/s00761-023-01351-8
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00761-023-01351-8