Volumetric Measurement of Tumors David F. Yankelevitz, MD
Why Measure Tumor Volumes? Surrogate for knowing the amount of viable tumor Implied is this: –Larger volumes, therefore progression –Smaller volumes, therefore response
How do we measure volumes? Surrogates –Uni-dimension (RECIST) –Bi-dimension (WHO) –Tri-dimension Genuine volume measurements
Advantages of Volume Measurements Greater proportional change –26% diameter increase corresponds to 100% volume increase Measurement of asymmetric growth Tumor volume doubling time
Days to ERCT from initial CT Initial nodule diameter (mm) Doubling time (days) (24 vs 93%) (4 vs 12%) Expected Change in Diameter
Asymmetric Growth SPN (6.9 mm) at baseline and 36 days later Virtually unchanged according to 2D metrics Apparently benign (DT=9700) Area: 36.5 mm 2 Perimeter: 22.7 mm Length: 8.27 mm Width: 5.62 mm Area: 36.6 mm 2 Perimeter: 23.4 mm Length: 8.23 mm Width: 5.66 mm
Volumetric Analysis 3D analysis reveals significant growth along scanner axis! (DT = 104, malignant)
8 mm Stable Nodule
Volumetric Growth Rate Analysis 8 mm stable pulmonary nodule at baseline and 181 days later MVGI = 0.57%
10 mm Malignant Nodule
Volumetric Growth Rate Analysis 10 mm malignant pulmonary nodule at baseline and 32 days later MVGI = 22.0% -- Squamous Cell Carcinoma
Inputs Into Volume Estimates Accuracy of measuring device (machine) –Inplane (x,y), out of plane (z) Ability to define borders of target (anatomic) –Removal of attached structures CAD –Defining edges Margin of tumor Adjacent edema/inflammation –Stability of structures
10mm Slice Thickness (Anisotropic) © ELCAP 2002
5mm Slice Thickness © ELCAP 2002
2.5mm Slice Thickness © ELCAP 2002
1mm Slice Thickness © ELCAP 2002
Headline Courtesy of University of Erlangen, Department of Radiology and Institute of Medical Physics SOMATOM Sensation 64 6 sec for 400 mm 64 x 0.6mm (2x32) Resolution 0.4 mm Rotation 0.37 sec 120 kV / 100 mAs
Headline Courtesy of University of Erlangen, Department of Radiology and Institute of Medical Physics SOMATOM Sensation 64 6 sec for 400 mm 64 x 0.6mm (2x32) Resolution 0.4 mm Rotation 0.37 sec 120 kV / 100 mAs
Volumetric CT Scanning
Accuracy of Area Measurements
Deformable Synthetic Nodules
Volumetric Measurement - Synthetic Nodules Volume Error:(3-6 mm) = 1.1% RMS, 2.8% max (6-11 mm) = 0.5% RMS, 0.9% max Function of nodule size Yankelevitz, et al. Radiology 2000
Removal of Attached Structures Jan , (X,Y) resolution: mm, Slice thickness : 1 mm Images ©1999, ELCAP Lab, Weill Medical College of Cornell University
Solid Nodule Segmentation
Images ©1999, ELCAP Lab, Weill Medical College of Cornell University 74-Day Doubling Time Volumetric Doubling Time Estimation
Limitations of Segmentation
Part-Solid Nodule: Complex Segmentation
Abutting Pleura: Limitations of Segmentation
Nonsolid Nodule: Indistinct Border
Less Natural Contrast
SPICULATED NODULE Instructions to Thoracic Radiologists were “Draw the Boundary of the Nodule”
SPICULATED NODULE Expert Number 1 Contour
SPICULATED NODULE Expert Number 2 Contour
SPICULATED NODULE Comparison of Contours
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Patient Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Nonsolid nodule: Adenocarcinoma, bronchioloalveolar subtype
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Motion Artifact – Cardiac Motion Images © , ELCAP Lab, Weill Medical College of Cornell University
Nodule Growth Rates Exponential Growth Model Nodule Doubling Time (DT) Traditional 2D Approximation
Appropriate Time to Follow-up CT When should the follow-up CT be done? where d is the reliably-detectable percent volume change, a function of initial nodule size d = two standard deviations of PVC in stable nodules by size category DT D = 400 days for baseline cases DT D = upper bound on doubling time for repeat cases example: 208 days for 3 mm nodule
Time to Follow-up CT Appropriate time to follow-up CT by initial nodule size detected on baseline or repeat screening Time to Follow-up CT (days) for nodules detected on Size (mm) (d) (%)BaselineRepeat
Review of Literature Limited data on comparison of 3D volume measurements to 2D or 1D, notably for large lesions Most report that volume is better for large ‘well-defined’ abnormalities Limited impact on change in category for RECIST
Summary Technology has greatly improved –Measuring device –Image processing Little work has been done in regard to complex abnormalities Potential to markedly improve response estimates
Volumetric Measurement of In Vivo Nodules Although we had quantified the relative error in phantom nodule measurement by size, the error for in vivo nodules must be greater –partial volume –vascular geometry –motion artifacts
Assessment of In Vivo Volume Estimates Rescanning in short interval –Smallest change in true nodule volume –Difficult study due to dose concerns Stable nodules –Scans more easily obtained (screening) –Accounts for small errors in patient positioning and scanner calibration drift
Cases 262 HRCT scans of 120 stable nodules –Standard dose, small FOV, HRCT –Nodules 2-11 mm in diameter –Determination of stability based on radiologist evaluation over period of 2 or more years –Assessment of technical artifacts Incomplete acquisition System error –Assessment of motion artifacts Five-point scale Patient motion (gross movement, respiration) Cardiac motion 20 HRCT scans of 10 malignant nodules
Artifacts Motion artifact and technical artifact in 262 CT scans by initial nodule size Technical Motion Artifact ScoreAny Artifact NoneMinimalModeratePronouncedSevere Artifact Size (mm) Any size (22%) of cases had to be excluded due to technical or motion artifacts
Stable Nodules Frequency distribution of 94 stable nodules by initial size and time to follow-up CT Time to Follow-up CT (months) Any Interval Size (mm) Any size
Monthly Volumetric Growth Index Monthly Volumetric Growth Index, MVGI –Percent change in volume per month –Remaps growth estimates into two distinct classes Reeves, et al. RSNA 2001
Nodule Growth Rates MVGI = 32.5 MVGI = 15.1 MVGI = 4.3
MVGI of Stable Nodules Mean and standard deviation of monthly volumetric growth index of 94 stable nodules by initial size and time to follow-up CT Time to Follow-up CT (months) Any Interval Size (mm) MeanSDMeanSDMeanSDMeanSD Any Size Overall mean 0.06% Std. Err. of the Mean 0.21%
MVGI of Stable Nodules Mean and standard deviation of monthly volumetric growth index of 94 stable nodules by initial size and time to follow-up CT Time to Follow-up CT (months) Any Interval Size (mm) MeanSDMeanSDMeanSDMeanSD Any Size SD decreases with increasing size SD decreases with increasing time to follow-up CT
PVC of Stable Nodules Mean and standard deviation of percent volume change of 94 stable nodules by initial size and time to follow-up CT Time to Follow-up CT (months) Any Interval Size (mm) MeanSDMeanSDMeanSDMeanSD Any Size
PVC of Stable Nodules Mean and standard deviation of percent volume change of 94 stable nodules by initial size and time to follow-up CT Time to Follow-up CT (months) Any Interval Size (mm) MeanSDMeanSDMeanSDMeanSD Any Size SD decreases with increasing size SD increases with increasing time to follow-up CT
Malignant Nodules Monthly volumetric growth index of 10 malignant nodules with initial size, time to follow-up CT, and histologic diagnosis Initial Time to Follow-upHistologic CaseDetectionSize (mm)CT (days)MVGI (%)Diagnosis 1 Baseline Adenocarcinoma 2 Baseline Squamous Cell 3 Baseline Large Cell 4 Baseline Adenocarcinoma 5 Repeat Adenocarcinoma 6 Repeat Squamous Cell 7 Repeat Adenocarcinoma 8 Repeat Adenocarcinoma 9 Repeat Adenocarcinoma 10 Repeat Large Cell
Comparison of MVGI Values All of the stable nodules had values within two standard deviations of the corresponding mean value by size, while each of the 10 malignant nodules exceeded that corresponding value.
Conclusions The mean value of MVGI for stable nodules was 0.06% and its standard error was 0.21%. All of the stable nodules had values within two standard deviations of the corresponding mean value by size, while each of the 10 malignant nodules exceeded that corresponding value. Conclusion: Three-dimensional computer methods can be used to reliably characterize growth in small solid pulmonary nodules. Factors affecting the reproducibility of growth rate estimates include the initial nodule size, the timing of the follow-up scan, and the presence of patient-induced or technical artifacts.