S2469
Physics - Radiomics, functional and biological imaging, and outcome prediction
ESTRO 2026
Physical, and Laboratory Examinations. 3rd edition. Boston: Butterworths; 1990. Chapter 151. Available from: https://www.ncbi.nlm.nih.gov/books/NBK259/3. Tofts, Paul S. “Modeling tracer kinetics in dynamic Gd- DTPA MR imaging.” Journal of magnetic resonance imaging : JMRI vol. 7,1 (1997): 91-101 Keywords: DCE-MRI, vascular input function, perfusion Digital Poster Highlight 3604 Can we use simulated contours to assess the reproducibility of radiomic features to interobserver contour variation? Rhianna Brown 1,2 , Amy Walker 2,3 , Karen Lim 4,5 , Dean Cutajar 1 , Peter Metcalfe 1 , Lois Holloway 2,3 1 Centre for Medical Radiation Physics, University of Wollongong, Wollongong, Australia. 2 Department of Medical Physics, Liverpool and Macarthur Cancer Therapy Centre, Liverpool, Australia. 3 Medical Physics, 2. Ingham Institute for Applied Medical Research, Liverpool, Australia. 4 South Western Sydney Local Health District, Liverpool and Macarthur Cancer Therapy Centre, Liverpool, Australia. 5 South Western Sydney Clinical School, University of New South Wales, Liverpool, Australia Purpose/Objective: The reproducibility of radiomic features to interobserver contour variation (IOV) is typically assessed using ≤ 4 observers. This has been shown to not always ensure a radiomic feature’s reproducibility to IOV [1]. However, obtaining additional observer’s contours is difficult due to the time intensive nature of contouring. This study assessed the reproducibility of radiomic features extracted from contours generated using two different contour simulation methods compared to real observers’ contours. Material/Methods: 20 gynaecological cancer T2W-MRIs were delineated by six clinicians. Two methods were used to simulate the observer contours. One approach used the delineation error [2] of the available observer contours to determine a margin to generate estimated maximum and minimum contours. Between these contours, a singular contour is geometrically varied to obtain simulated contours (max/min method). The other method used an approach similar to that described by Osorio et al. [3], however, larger noise scales were required to adequately generate simulated contours for gynaecological IOV. From the observers’ and simulated contours, 107 radiomic features were extracted using PyRadiomics, with the reproducibility of these features assessed using an intraclass correlation coefficient (ICC). Excellent, good, moderate and poor reproducibility was determine
Figure 2: Patient-wise relative change in Ktrans between the two timepoints when varying blood T1, hematocrit and time shift. A reversal in trend corresponds to at least one value of opposite sign than the reference. Conclusion: Utilizing different parameter choices between the timepoints for Ktrans showed that results could be reversed with respect to the reference trend. Across the tested parameter ranges, varying T1 showed the largest relative change of Ktrans. Moreover, varying hematocrit had a larger effect than time shifting of the VIF. References: 1. Stanisz, Greg J., et al. "T1, T2 relaxation and magnetization transfer in tissue at 3T." Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine 54.3 (2005): 507-512.2. Billett H. H. “Hemoglobin and Hematocrit”. In: Walker H. K., Hall W. D., Hurst J. W., editors. Clinical Methods: The History,
Made with FlippingBook - Share PDF online