Laura Close Medical Biophysics The University of Western Ontario March 22, 2011

1. INTRA-OBSERVER REPRODUCIBILITY AND ACCURACY OF 1D, 2D, AND 3D LUNG TUMOUR MEASUREMENTS Laura Close Medical Biophysics The University of Western Ontario March 22, 2011

2. Acknowledgements • Dr. Grace Parraga • Mr. Amir Owrangi • Ms. Lauren Villemaire • Mr. Andrew Wheatley

3. Introduction • Lung cancer is the leading cause cancer-related deaths in Canada1 • Many variables contributing to effectiveness of patient treatment • Choice/length of treatment • Monitoring of treatment • tracking tumour size • Need effective measuring technique 1The future of cancer control in Canada. (2011). Canadian Partnership Against Cancer.

4. Theory • 1979: World Health Organization (WHO) 1 • 2D measurement • (longest diameter) x (longest perpendicular bisector) • 2000: Response Evaluation Criteria in Solid Tumours (RECIST) 1 • 1D measurement • Longest diameter • 1D & 2D Limitations: • Ex: Out-of-plane dimensions, sphericity assumed1 • 3D Quantify, created at Robarts Research Institute (London, ON) • Volume measurement • Contours & 3D triangular meshes • Irregular shapes, asymmetrical growth rates1 1Wilson, L.C.R. (2010). Development of multi-dimensional x-ray computed tomography measurements of lung tumours (Master’s thesis).

5. Objectives • Determine intra-observer reproducibility of 1D, 2D, and 3D measurement techniques • Determine accuracy of 1D, 2D, and 3D measurements in relation to ground truth measurements • Use these factors to gain insight into whether 3D measurements are appropriate for clinical settings

6. Approach • X-ray CT images: • 2 patient tumours at 9 time points • 3 phantom tumours at 4 slice thicknesses • Software programs: • ClearCanvas • 3D Quantify • Measurements: • 1D: RECIST (ClearCanvas) • 2D: WHO (ClearCanvas) • 3D: Volume (3D Quantify)

7. Methods • Ensuring Non-Bias: • 2 patient tumours x 9 time points x 5 rounds & • 3 phantoms x 4 slice thicknesses x 5 rounds • All randomized • Blind to ground truth measurements for phantoms • Did not exceed 1 round of measurements per day

8. 1D Measurements(RECIST Diameter) • Using ClearCanvas • Longest diameter Patient Tumour (Large), Time Point 1 Patient Tumour (Small), Time Point 1 Phantom Tumour (Medium), 0.5mm Slice Thickness Phantom Tumour (Small), 0.5mm Slice Thickness Phantom Tumour (Large), 0.5mm Slice Thickness

9. 2D Measurements(WHO) • Using ClearCanvas • Longest diameter x longest perpendicular bisector Patient Tumour (Large), Time Point 1 Patient Tumour (Small), Time Point 1 Phantom Tumour (Medium), 0.5mm Slice Thickness Phantom Tumour (Small), 0.5mm Slice Thickness Phantom Tumour (Large), 0.5mm Slice Thickness

10. 3D Measurements(Volume) • Using 3D Quantify • Volume of triangular mesh Patient Tumour (Large), Time Point 1 Patient Tumour (Small), Time Point 1 Phantom Tumour (Medium), 0.5mm Slice Thickness Phantom Tumour (Small), 0.5mm Slice Thickness Phantom Tumour (Large), 0.5mm Slice Thickness

11. Ex: Generation of Medium Phantom 3D Mesh

12. Results Patient Tumour Size Over Time (Large)

13. Patient Tumour Size Over Time (Small)

14. Measured Phantom Size Vs. Slice Thickness No obvious trends in relation to slice thickness 0.5mm Slice Thickness 5.0mm Slice Thickness Large Large Large Medium Medium Small Medium Small Small

15. Increased slice thickness relates to an increase in volume measurement Measured Phantom Size Vs. Slice Thickness 0.5mm Slice Thickness 5.0mm Slice Thickness Large Large Medium Medium Small Small

16. Intra-Observer Reproducibility: Intraclass Correlation Coefficients Phantom Tumours • For clinical measurements, ICC values ≥0.9 recommended1 (Indicated in green) • 1D, 2D, and 3D phantom tumour measurements are reliable • 1D, 2D, and some 3D patient tumour measurements are reliable • Certain time points more obscured Slice Thickness Patient Tumours Time Point Small and large tumours, time point 9 1 Portney LG, Watkins MP. Foundations of clinical research: applications to practice. Norwalk, CT: Appleton & Lange 1993; 505-528.

17. Accuracy of Measurements in Relation to Known Ground Truth Measurements • T-tests compare image measurements to ground truth measurements (for each tumour at each slice thickness) • 2-sample, 2-tailed t-tests, assuming unequal variance (all F-test p-values were <0.05) • H0: Means are equal • Using α=0.05, H0 is not rejected for values in green • 3D measurements show strongest potential for reproducing ground truths

18. Discussion • Results account for: imaging, software, measurement technique, individual • Measurement technique of great importance • 3D volume measurement favourable • 3D measurements are reproducible • 3D recreates ground truth measurements • Ability to measure irregular phantoms promising as real tumours usually demonstrate simpler geometry

19. Discussion (cont.) • Tumour measurements impact research/development of treatments • Roadblocks for 3D measurements: • Imaging resolution limitations impacting 3D accuracy overcome • Time-consuming • Could stress time is worthwhile for truer measurements • Could increase efficiency of 3D method Vs.

20. Conclusion • 1D, 2D, and 3D measurements are reproducible • 3D measurements display greatest potential in accurately recreating ground truth measurements • Definite advantages in clinical settings once drawback of time-consumption is overcome