Radiation injury vs. recurrent brain metastasis: combining textural feature radiomics analysis and standard parameters may increase 18F-FET PET accuracy without dynamic scans
- 584 Downloads
We investigated the potential of textural feature analysis of O-(2-[18F]fluoroethyl)-L-tyrosine (18F-FET) PET to differentiate radiation injury from brain metastasis recurrence.
Forty-seven patients with contrast-enhancing brain lesions (n = 54) on MRI after radiotherapy of brain metastases underwent dynamic 18F-FET PET. Tumour-to-brain ratios (TBRs) of 18F-FET uptake and 62 textural parameters were determined on summed images 20-40 min post-injection. Tracer uptake kinetics, i.e., time-to-peak (TTP) and patterns of time-activity curves (TAC) were evaluated on dynamic PET data from 0-50 min post-injection. Diagnostic accuracy of investigated parameters and combinations thereof to discriminate between brain metastasis recurrence and radiation injury was compared.
Diagnostic accuracy increased from 81 % for TBRmean alone to 85 % when combined with the textural parameter Coarseness or Short-zone emphasis. The accuracy of TBRmax alone was 83 % and increased to 85 % after combination with the textural parameters Coarseness, Short-zone emphasis, or Correlation. Analysis of TACs resulted in an accuracy of 70 % for kinetic pattern alone and increased to 83 % when combined with TBRmax.
Textural feature analysis in combination with TBRs may have the potential to increase diagnostic accuracy for discrimination between brain metastasis recurrence and radiation injury, without the need for dynamic 18F-FET PET scans.
• Textural feature analysis provides quantitative information about tumour heterogeneity
• Textural features help improve discrimination between brain metastasis recurrence and radiation injury
• Textural features might be helpful to further understand tumour heterogeneity
• Analysis does not require a more time consuming dynamic PET acquisition
KeywordsFET PET Textural analysis Radiomics Brain metastasis Radiation injury
Area under the receiver-operating-characteristic curve
Full width at half maximum
Magnetic resonance spectroscopy
Normalized co-occurrence matrix
Neighbourhood grey level dependence
Neighbourhood intensity difference matrix
Ordered subset expectation maximisation
Standardized uptake value
Mean tumour-to-brain ratio
Maximum tumour-to-brain ratio
Texture feature coding
Texture feature coding co-occurrence matrix
Whole-brain radiation therapy
The authors thank Suzanne Schaden, Elisabeth Theelen, Silke Frensch, Kornelia Frey and Lutz Tellmann for assistance in the patient studies; Johannes Ermert, Silke Grafmüller, Erika Wabbals and Sascha Rehbein for radiosynthesis of 18F-FET.
The scientific guarantor of this publication is Prof. Dr. Karl-Josef Langen. The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article. The authors state that this work has not received any funding.
No complex statistical methods were necessary for this paper. Institutional Review Board approval was obtained. Written informed consent was obtained from all subjects (patients) in this study.
Some study subjects or cohorts have been previously reported in Galldiks N, Stoffels G, Filss CP et al (2012) Role of O-(2-(18)F-fluoroethyl)-L-tyrosine PET for differentiation of local recurrent brain metastasis from radiation necrosis. J Nucl Med 53:1367–1374.
Methodology: retrospective, diagnostic study, performed at one institution.
- 33.Marusyk A, Polyak K (2011) Tumor heterogeneity: causes and consequences. Biochim Biophys Acta 1805:1–28Google Scholar
- 39.Murrell DH, Hamilton AM, Mallett CL et al (2015) Understanding Heterogeneity and Permeability of Brain Metastases in Murine Models of HER2-Positive Breast Cancer Through Magnetic Resonance Imaging: Implications for Detection and Therapy. Transl Oncol 8:176–184CrossRefPubMedPubMedCentralGoogle Scholar
- 46.Haralick RM, Shanmugam K, Dinstein I (1973) Textural Features for Image Classification. IEEE Trans Syst Man Cybern 3:610–621Google Scholar
- 49.Thibault G, Fertil B, Navarro C, et al (2009) Texture Indexes and Gray Level Size Zone Matrix Application to Cell Nuclei Classification. Pattern Recognit Inf Process 140–145.Google Scholar
- 54.Geertzen J (2012) Inter-Rater Agreement with multiple raters and variables. https:/nlp-ml.io/jg/software/ira/. Accessed 27 Apr 2016Google Scholar
- 55.Lowry R (1998) VassarStats: Website for Statistical Computation. http://vassarstats.net/roc_comp.html. Accessed 29 Sep 2016Google Scholar
- 61.Bailly C, Bodet-Milin C, Couespel S et al (2016) Revisiting the robustness of PET-based textural features in the context of multi-centric trials. PLoS One 11:1–16Google Scholar
- 62.Piroth MD, Liebenstund S, Galldiks N et al (2013) Monitoring of radiochemotherapy in patients with glioblastoma using O-(2-[18F]fluoroethyl)-L-tyrosine positron emission tomography: Is dynamic imaging helpful? Mol Imaging 12:1–8Google Scholar