Die Wichtigkeit des „Machine Learning“ und der Texturanalyse in der Onkologie anhand der Fusionsbildgebung

Zusammenfassung

Das Gebiet der Onkologie hat sich in den letzten Jahrzehnten deutlich weiterentwickelt. Klassische Chemotherapien werden aktuell zunehmend durch neue, zielgerichtete „State-of-the-Art“-Therapien ersetzt. Diese zielgerichteten Therapien basieren auf einer Zusammenführung von Informationen aus der molekularen Pathologie, spezifischen Tumormarkern und fortgeschrittener Bildgebungstechnik. PET/CT und PET/MRT bieten aufseiten der Bildgebung hierbei die einzigartige Möglichkeit funktionelle und anatomische Informationen in einer Untersuchung zu vereinen. Bis vor Kurzem wurden PET- und CT/MRT-Bilder meistens binär (Tumor ja/nein) analysiert ohne wesentliche Informationen über die räumliche Verteilung von relevanten Bildparametern innerhalb einer Tumorläsion zu erheben. Mit der Entwicklung und dem Voranschreiten von Texturanalyse-Softwares der nächsten Generation ist es jetzt möglich eine Tumorläsion auch bez. ihrer Heterogenität und Expression spezifischer Zielstrukturen auf den Tumorzellen zu charakterisieren. Jedoch gibt es bei diesem Vorgehen mehrere Herausforderungen, wobei die wichtigste ist, die durch die Texturanalysen generierten großen Datenmengen zu interpretieren. Der Erfolg des „Machine Learning“/des maschinellen Lernens in anderen Gebieten der Wissenschaft, welche mit ähnlich großen Datenmengen umgehen müssen, hat ein neues Kapitel in der zielgerichteten Onkologie mit radioaktiv-markierten Proben eröffnet. In unserem Review-Artikel haben wir das primäre Ziel, aktuelle Literatur und zukünftige Trends der Integration von „Machine Learning“ in der Onkologie durch die Nutzung der Fusionsbildgebung mittels PET/CT oder PET/MRT zusammengefasst.

Abstract

The field of oncology has advanced significantly in the last couple of decades. „Blanket treatment“ with chemotherapy is now being slowly replaced by state of the art targeted therapy. These targeted therapies are guided by integration of information from molecular pathology, surrogate tumor markers, and advanced imaging technologies. PET/CT and PET/MRI offer the unique opportunity to combine functional and anatomical information. Until recently the PET and CT/MRI images were mostly analyzed as „structural binary changes“ with little information on the spatial distribution of relevant imaging parameters within a tumor lesion. With advancement of next generation texture analyses software it is now possible to characterize a tumor lesion with respect to its heterogeneity and target expression. However, there are several challenges in this approach most important amongst them is the ways and means of interpreting vast data generated through the texture analyses. Success of machine learning in other fields of science dealing with similarly large data has opened up a new chapter in precision oncology with radiolabeled probes. In our review article we have summarized the primary aim, current literature and future trends in the integration of machine learning in oncology using PET/CT or PET/MRI images.

Schlüsselwörter

„Machine Learning“ / maschinelles Lernen - Radiomics - Fusionsbildgebung - FDG-PET/CT - Texturanalyse

Keywords

Machine Learning - Big Data - hybrid imaging - PET - texture analysis

• References

• 1 Blanc-Durand P, van der Gucht A, Schaefer N. et al. Automatic lesion detection and segmentation of 18F-FET PET in gliomas: A full 3D U-Net convolutional neural network study. PLoS One 2018; 13
• 2 Bleeker G, Tytgat GAM, Adam JA. et al. 123I-MIBG scintigraphy and 18F-FDG-PET imaging for diagnosing neuroblastoma. Cochrane Database Syst Rev 2015; CD009263
• 3 Buizza G, Toma-Dasu I, Lazzeroni M. et al. Early tumor response prediction for lung cancer patients using novel longitudinal pattern features from sequential PET/CT image scans. Phys Med 2018; 54: 21-29
• 4 Castellano G, Bonilha L, Li LM. et al. Texture analysis of medical images. Clin Radiol 2004; 59: 1061-1069
• 5 Cistaro A, Quartuccio N, Caobelli F. et al. 124I-MIBG: a new promising positron-emitting radiopharmaceutical for the evaluation of neuroblastoma. Nucl Med Rev Cent East Eur 2015; 18: 102-106
• 6 Cook GJR, Yip C, Siddique M. et al. Are pretreatment 18F-FDG PET tumor textural features in non-small cell lung cancer associated with response and survival after chemoradiotherapy?. J Nucl Med 2013; 54: 19-26
• 7 Cook GJR, Azad G, Owczarczyk K. et al. Challenges and Promises of PET Radiomics. Int J Radiat Oncol Biol Phys 2018; 102: 1083-1089
• 8 Feng M, Valdes G, Dixit N. et al. Machine Learning in Radiation Oncology: Opportunities, Requirements, and Needs. Front Oncol 2018; 8: 110
• 9 Kebir S, Weber M, Lazaridis L. et al. Hybrid 11C-MET PET/MRI Combined With "Machine Learning" in Glioma Diagnosis According to the Revised Glioma WHO Classification 2016. Clin Nucl Med 2019; 44: 214-220
• 10 Kiratli PÖ, Tuncel M, Bar-Sever Z. Nuclear Medicine in Pediatric and Adolescent Tumors. Semin Nucl Med 2016; 46: 308-323
• 11 Li H, Galperin-Aizenberg M, Pryma D. et al. Unsupervised machine learning of radiomic features for predicting treatment response and overall survival of early stage non-small cell lung cancer patients treated with stereotactic body radiation therapy. Radiother Oncol 2018; 129: 218-226
• 12 Melzer HI, Coppenrath E, Schmid I. et al. ¹²³I-MIBG scintigraphy/SPECT versus ¹⁸F-FDG PET in paediatric neuroblastoma. Eur J Nucl Med Mol Imaging 2011; 38: 1648-1658
• 13 Nakagawa M, Nakaura T, Namimoto T. et al. A multiparametric MRI-based machine learning to distinguish between uterine sarcoma and benign leiomyoma: comparison with 18F-FDG PET/CT. Clin Radiol 2019; 74: 167.e1-167.e7
• 14 Neubauer H, Li M, Müller VR. et al. Diagnostischer Stellenwert der diffusionsgewichteten MRT zur Tumorcharakterisierung, Tumordifferenzierung und zur Verlaufskontrolle bei pädiatrischen Patienten mit neuroblastischen Tumoren. Rofo 2017; 189: 640-650
• 15 Palmerini E, Colangeli M, Nanni C. et al. The role of FDG PET/CT in patients treated with neoadjuvant chemotherapy for localized bone sarcomas. Eur J Nucl Med Mol Imaging 2016; 44: 215-223
• 16 Pandit-Taskar N, Zanzonico P, Staton KD. et al. Biodistribution and Dosimetry of 18F-Meta-Fluorobenzylguanidine: A First-in-Human PET/CT Imaging Study of Patients with Neuroendocrine Malignancies. J Nucl Med 2018; 59: 147-153
• 17 Papathanasiou ND, Gaze MN, Sullivan K. et al. 18F-FDG PET/CT and 123I-metaiodobenzylguanidine imaging in high-risk neuroblastoma: diagnostic comparison and survival analysis. J Nucl Med 2011; 52: 519-525
• 18 Papp L, Pötsch N, Grahovac M. et al. Glioma Survival Prediction with Combined Analysis of In Vivo 11C-MET PET Features, Ex Vivo Features, and Patient Features by Supervised Machine Learning. J Nucl Med 2018; 59: 892-899
• 19 Peeken JC, Goldberg T, Pyka T. et al. Combining multimodal imaging and treatment features improves machine learning‐based prognostic assessment in patients with glioblastoma multiforme. Cancer Med 2018; 8: 128-136
• 20 Perk T, Bradshaw T, Chen S. et al. Automated classification of benign and malignant lesions in 18F-NaF PET/CT images using machine learning. Phys Med Biol 2018; 63: 225019
• 21 Piccardo A, Puntoni M, Lopci E. et al. Prognostic value of ¹⁸F-DOPA PET/CT at the time of recurrence in patients affected by neuroblastoma. Eur J Nucl Med Mol Imaging 2014; 41: 1046-1056
• 22 Sagiyama K, Watanabe Y, Kamei R. et al. Multiparametric voxel-based analyses of standardized uptake values and apparent diffusion coefficients of soft-tissue tumours with a positron emission tomography/magnetic resonance system: Preliminary results. Eur Radiol 2017; 27: 5024-5033
• 23 Sharp SE, Shulkin BL, Gelfand MJ. et al. 123I-MIBG scintigraphy and 18F-FDG PET in neuroblastoma. J Nucl Med 2009; 50: 1237-1243
• 24 Simon T, Hero B, Schulte JH. et al. 2017 GPOH Empfehlungen für Diagnostik und Therapie von Patienten mit neuroblastischen Tumoren. Klin Padiatr 2017; 229: 147-167
• 25 Surov A, Meyer HJ, Wienke A. Standardized Uptake Values Derived from 18F-FDG PET May Predict Lung Cancer Microvessel Density and Expression of KI 67, VEGF, and HIF-1α but Not Expression of Cyclin D1, PCNA, EGFR, PD L1, and p53. Contrast Media Mol Imaging 2018; 2018: 1-10
• 26 Takeda K, Takanami K, Shirata Y. et al. Clinical utility of texture analysis of 18F-FDG PET/CT in patients with Stage I lung cancer treated with stereotactic body radiotherapy. J Radiat Res 2017; 58: 862-869
• 27 Tsujikawa T, Yamamoto M, Shono K. et al. Assessment of intratumor heterogeneity in mesenchymal uterine tumor by an 18F-FDG PET/CT texture analysis. Ann Nucl Med 2017; 31: 752-757
• 28 Vaidya M, Creach KM, Frye J. et al. Combined PET/CT image characteristics for radiotherapy tumor response in lung cancer. Radiother Oncol 2012; 102: 239-245
• 29 Vallières M, Freeman CR, Skamene SR. et al. A radiomics model from joint FDG-PET and MRI texture features for the prediction of lung metastases in soft-tissue sarcomas of the extremities. Phys Med Biol 2015; 60: 5471-5496
• 30 Wang H, Zhou Z, Li Y. et al. Comparison of machine learning methods for classifying mediastinal lymph node metastasis of non-small cell lung cancer from 18F-FDG PET/CT images. EJNMMI Res 2017; 7: 1-11
• 31 Wolsztynski E, O'Sullivan F, Keyes E. et al. Positron emission tomography-based assessment of metabolic gradient and other prognostic features in sarcoma. J Med Imaging (Bellingham) 2018; 5: 24502
• 32 Yu H, Caldwell C, Mah K. et al. Coregistered FDG PET/CT-based textural characterization of head and neck cancer for radiation treatment planning. IEEE Trans Med Imaging 2009; 28: 374-383