Evaluation of a commercial DIR platform for contour propagation in prostate cancer patients treated with IMRT/VMAT.

PURPOSE: To assess the performance and limitations of contour propagation with three commercial deformable image registration (DIR) algorithms using fractional scans of CT-on-rails (CTOR) and Cone Beam CT (CBCT) in image guided prostate therapy patients treated with IMRT/VMAT. METHODS: Twenty prostate cancer patients treated with IMRT/VMAT were selected for analysis. A total of 453 fractions across those patients were analyzed. Image data were imported into MIM (MIM Software, Inc., Cleveland, OH) and three DIR algorithms (DIR Profile, normalized intensity-based (NIB) and shadowed NIB DIR algorithms) were applied to deformably register each fraction with the planning CT. Manually drawn contours of bladder and rectum were utilized for comparison against the DIR propagated contours in each fraction. Four metrics were utilized in the evaluation of contour similarity, the Hausdorff Distance (HD), Mean Distance to Agreement (MDA), Dice Similarity Coefficient (DSC), and Jaccard indices. A subfactor analysis was performed per modality (CTOR vs. CBCT) and time (fraction). Point estimates and 95% confidence intervals were assessed via a Linear Mixed Effect model for the contour similarity metrics. RESULTS: No statistically significant differences were observed between the DIR Profile and NIB algorithms. However, statistically significant differences were observed between the shadowed NIB and NIB algorithms for some of the DIR evaluation metrics. The Hausdorff Distance calculation showed the NIB propagated contours vs. shadowed NIB propagated contours against the manual contours were 14.82 mm vs. 8.34 mm for bladder and 15.87 mm vs. 11 mm for rectum, respectively. Similarly, the Mean Distance to Agreement calculation comparing the NIB propagated contours vs. shadowed NIB propagated contours against the manual contours were 2.43 mm vs. 0.98 mm for bladder and 2.57 mm vs. 1.00 mm for rectum, respectively. The Dice Similarity Coefficients comparing the NIB propagated contours and shadowed NIB propagated contours against the manual contours were 0.844 against 0.936 for bladder and 0.772 against 0.907 for rectum, respectively. The Jaccard indices comparing the NIB propagated contours and shadowed NIB propagated contours against the manual contours were 0.749 against 0.884 for bladder and 0.637 against 0.831 for rectum, respectively. The shadowed NIB DIR, which showed the closest agreement with the manual contours performed significantly better than the DIR Profile in all the comparisons. The OAR with the greatest agreement varied substantially across patients and image guided radiation therapy (IGRT) modality. Intra-patient variability of contour metric evaluation was insignificant across all the DIR algorithms. Statistical significance at α = 0.05 was observed for manual vs. deformably propagated contours for bladder for all the metrics except Hausdorff Distance (P = 0.01 for MDA, P = 0.02 for DSC, P = 0.01 for Jaccard), whereas the corresponding values for rectum were: P = 0.03 for HD, P = 0.01 for MDA, P < 0.01 for DSC, P < 0.01 for Jaccard. The performance of the different metrics varied slightly across the fractions of each patient, which indicates that weekly contour propagation models provide a reasonable approximation of the daily contour propagation models. CONCLUSION: The high variance of Hausdorff Distance across all automated methods for bladder indicates widely variable agreement across fractions for all patients. Lower variance across all modalities, methods, and metrics were observed for rectum. The shadowed NIB propagated contours were substantially more similar to the manual contours than the DIR Profile or NIB contours for both the CTOR and CBCT imaging modalities. The relationship of each algorithm to similarity with manual contours is consistent across all observed metrics and organs. Screening of image guidance for substantial differences in bladder and rectal filling compared with the planning CT reference could aid in identifying fractions for which automated DIR would prove insufficient.

Towards a universal MRI atlas of the prostate and prostate zones : Comparison of MRI vendor and image acquisition parameters.

BACKGROUND AND PURPOSE: The aim of this study was to evaluate an automatic multi-atlas-based segmentation method for generating prostate, peripheral (PZ), and transition zone (TZ) contours on MRIs with and without fat saturation (±FS), and compare MRIs from different vendor MRI systems. METHODS: T2-weighted (T2) and fat-saturated (T2FS) MRIs were acquired on 3T GE (GE, Waukesha, WI, USA) and Siemens (Erlangen, Germany) systems. Manual prostate and PZ contours were used to create atlas libraries. As a test MRI is entered, the procedure for atlas segmentation automatically identifies the atlas subjects that best match the test subject, followed by a normalized intensity-based free-form deformable registration. The contours are transformed to the test subject, and Dice similarity coefficients (DSC) and Hausdorff distances between atlas-generated and manual contours were used to assess performance. RESULTS: Three atlases were generated based on GE_T2 (n = 30), GE_T2FS (n = 30), and Siem_T2FS (n = 31). When test images matched the contrast and vendor of the atlas, DSCs of 0.81 and 0.83 for T2 ± FS were obtained (baseline performance). Atlases performed with higher accuracy when segmenting (i) T2FS vs. T2 images, likely due to a superior contrast between prostate vs. surrounding tissue; (ii) prostate vs. zonal anatomy; (iii) in the mid-gland vs. base and apex. Atlases performance declined when tested with images with differing contrast and MRI vendor. Conversely, combined atlases showed similar performance to baseline. CONCLUSION: The MRI atlas-based segmentation method achieved good results for prostate, PZ, and TZ compared to expert contoured volumes. Combined atlases performed similarly to matching atlas and scan type. The technique is fast, fully automatic, and implemented on commercially available clinical platform.

Validation of a deformable MRI to CT registration algorithm employing same day planning MRI for surrogate analysis.

PURPOSE: Validating deformable multimodality image registrations is challenging due to intrinsic differences in signal characteristics and their spatial intensity distributions. Evaluating multimodality registrations using these spatial intensity distributions is also complicated by the fact that these metrics are often employed in the registration optimization process. This work evaluates rigid and deformable image registrations of the prostate in between diagnostic-MRI and radiation treatment planning-CT by utilizing a planning-MRI after fiducial marker placement as a surrogate. The surrogate allows for the direct quantitative analysis that can be difficult in the multimodality domain. METHODS: For thirteen prostate patients, T2 images were acquired at two different time points, the first several weeks prior to planning (diagnostic-MRI) and the second on the same day as the planning-CT (planning-MRI). The diagnostic-MRI was deformed to the planning-CT utilizing a commercially available algorithm which synthesizes a deformable image registration (DIR) algorithm from local rigid registrations. The planning-MRI provided an independent surrogate for the planning-CT for assessing registration accuracy using image similarity metrics, including Pearson correlation and normalized mutual information (NMI). A local analysis was performed by looking only within the prostate, proximal seminal vesicles, penile bulb, and combined areas. RESULTS: The planning-MRI provided an excellent surrogate for the planning-CT with residual error in fiducial alignment between the two datasets being submillimeter, 0.78 mm. DIR was superior to the rigid registration in 11 of 13 cases demonstrating a 27.37% improvement in NMI (P < 0.009) within a regional area surrounding the prostate and associated critical organs. Pearson correlations showed similar results, demonstrating a 13.02% improvement (P < 0.013). CONCLUSION: By utilizing the planning-MRI as a surrogate for the planning-CT, an independent evaluation of registration accuracy is possible. This population provides an ideal testing ground for MRI to CT DIR by obviating the need for multimodality comparisons which are inherently more challenging.

Correlation between two methods of florbetapir PET quantitative analysis.

This study evaluated performance of a commercially available standardized software program for calculation of florbetapir PET standard uptake value ratios (SUVr) in comparison with an established research method. Florbetapir PET images for 183 subjects clinically diagnosed as cognitively normal (CN), mild cognitive impairment (MCI) or probable Alzheimer's disease (AD) (45 AD, 60 MCI, and 78 CN) were evaluated using two software processing algorithms. The research method uses a single florbetapir PET template generated by averaging both amyloid positive and amyloid negative registered brains together. The commercial software simultaneously optimizes the registration between the florbetapir PET images and three templates: amyloid negative, amyloid positive, and an average. Cortical average SUVr values were calculated across six predefined anatomic regions with respect to the whole cerebellum reference region. SUVr values were well correlated between the two methods (r2 = 0.98). The relationship between the methods computed from the regression analysis is: Commercial method SUVr = (0.9757*Research SUVr) + 0.0299. A previously defined cutoff SUVr of 1.1 for distinguishing amyloid positivity by the research method corresponded to 1.1 (95% CI = 1.098, 1.11) for the commercial method. This study suggests that the commercial method is comparable to the published research method of SUVr analysis for florbetapir PET images, thus facilitating the potential use of standardized quantitative approaches to PET amyloid imaging.

Diagnostic performance of an automated analysis software for the diagnosis of Alzheimer's dementia with 18F FDG PET.

The objective of this study was to assess the ability of a quantitative software-aided approach to improve the diagnostic accuracy of 18F FDG PET for Alzheimer's dementia over visual analysis alone. Twenty normal subjects (M:F-12:8; mean age 80.6 years) and twenty mild AD subjects (M:F-12:8; mean age 70.6 years) with 18F FDG PET scans were obtained from the ADNI database. Three blinded readers interpreted these PET images first using a visual qualitative approach and then using a quantitative software-aided approach. Images were classified on two five-point scales based on normal/abnormal (1-definitely normal; 5-definitely abnormal) and presence of AD (1-definitely not AD; 5-definitely AD). Diagnostic sensitivity, specificity, and accuracy for both approaches were compared based on the aforementioned scales. The sensitivity, specificity, and accuracy for the normal vs. abnormal readings of all readers combined were higher when comparing the software-aided vs. visual approach (sensitivity 0.93 vs. 0.83 P = 0.0466; specificity 0.85 vs. 0.60 P = 0.0005; accuracy 0.89 vs. 0.72 P<0.0001). The specificity and accuracy for absence vs. presence of AD of all readers combined were higher when comparing the software-aided vs. visual approach (specificity 0.90 vs. 0.70 P = 0.0008; accuracy 0.81 vs. 0.72 P = 0.0356). Sensitivities of the software-aided and visual approaches did not differ significantly (0.72 vs. 0.73 P = 0.74). The quantitative software-aided approach appears to improve the performance of 18F FDG PET for the diagnosis of mild AD. It may be helpful for experienced 18F FDG PET readers analyzing challenging cases.