Synthetic pulmonary perfusion images from 4DCT for functional avoidance using deep learning.

Purpose.To develop and evaluate the performance of a deep learning model to generate synthetic pulmonary perfusion images from clinical 4DCT images for patients undergoing radiotherapy for lung cancer.Methods. A clinical data set of 58 pre- and post-radiotherapy99mTc-labeled MAA-SPECT perfusion studies (32 patients) each with contemporaneous 4DCT studies was collected. Using the inhale and exhale phases of the 4DCT, a 3D-residual network was trained to create synthetic perfusion images utilizing the MAA-SPECT as ground truth. The training process was repeated for a 50-imaging study, five-fold validation with twenty model instances trained per fold. The highest performing model instance from each fold was selected for inference upon the eight-study test set. A manual lung segmentation was used to compute correlation metrics constrained to the voxels within the lungs. From the pre-treatment test cases (N = 5), 50th percentile contours of well-perfused lung were generated from both the clinical and synthetic perfusion images and the agreement was quantified.Results. Across the hold-out test set, our deep learning model predicted perfusion with a Spearman correlation coefficient of 0.70 (IQR: 0.61-0.76) and a Pearson correlation coefficient of 0.66 (IQR: 0.49-0.73). The agreement of the functional avoidance contour pairs was Dice of 0.803 (IQR: 0.750-0.810) and average surface distance of 5.92 mm (IQR: 5.68-7.55).Conclusion. We demonstrate that from 4DCT alone, a deep learning model can generate synthetic perfusion images with potential application in functional avoidance treatment planning.

A prospective study to validate pulmonary blood mass changes on non-contrast 4DCT in pulmonary embolism patients.

PURPOSE: Limited diagnostic options exist for patients with suspected pulmonary embolism (PE) who cannot undergo CT-angiogram (CTA). CT-ventilation methods recover respiratory motion-induced lung volume changes as a surrogate for ventilation. We recently demonstrated that pulmonary blood mass change, induced by tidal respiratory motion, is a potential surrogate for pulmonary perfusion. In this study, we examine blood mass and volume change in patients with PE and parenchymal lung abnormalities (PLA). METHODS: A cross-sectional analysis was conducted on a prospective, cohort-study with 129 consecutive PE suspected patients. Patients received 4DCT within 48 h of CTA and were classified as having PLA and/or PE. Global volume change (VC) and percent global pulmonary blood mass change (PBM) were calculated for each patient. Associations with disease type were evaluated using quantile regression. RESULTS: 68 of 129 patients were PE positive on CTA. Median change in PBM for PE-positive patients (0.056; 95% CI: 0.045, 0.068; IQR: 0.051) was smaller than that of PE-negative patients (0.077; 95% CI: 0.064, 0.089; IQR: 0.056), with an estimated difference of 0.021 (95% CI: 0.003, 0.038; p = 0.0190). PLA was detected in 57 (44.2%) patients. Median VC for PLA-positive patients (1.26; 95% CI: 1.22, 1.30; IQR: 0.15) showed no significant difference from PLA-negative VC (1.25; 95% CI: 1.21, 1.28; IQR: 0.15). CONCLUSIONS: We demonstrate that pulmonary blood mass change is significantly lower in PE-positive patients compared to PE-negative patients, indicating that PBM derived from dynamic non-contrast CT is a potentially useful surrogate for pulmonary perfusion.

Quantifying pulmonary perfusion from noncontrast computed tomography.

PURPOSE: Computed tomography (CT)-derived ventilation methods compute respiratory induced volume changes as a surrogate for pulmonary ventilation. Currently, there are no known methods to derive perfusion information from noncontrast CT. We introduce a novel CT-Perfusion (CT-P) method for computing the magnitude mass changes apparent on dynamic noncontrast CT as a surrogate for pulmonary perfusion. METHODS: CT-Perfusion is based on a mass conservation model which describes the unknown mass change as a linear combination of spatially corresponding inhale and exhale HU estimated voxel densities. CT-P requires a deformable image registration (DIR) between the inhale/exhale lung CT pair, a preprocessing lung volume segmentation, and an estimate for the Jacobian of the DIR transformation. Given this information, the CT-P image, which provides the magnitude mass change for each voxel within the lung volume, is formulated as the solution to a constrained linear least squares problem defined by a series of subregional mean magnitude mass change measurements. Similar to previous robust CT-ventilation methods, the amount of uncertainty in a subregional sample mean measurement is related to measurement resolution and can be characterized with respect to a tolerance parameter τ . Spatial Spearman correlation between single photon emission CT perfusion (SPECT-P) and the proposed CT-P method was assessed in two patient cohorts via a parameter sweep of τ . The first cohort was comprised of 15 patients diagnosed with pulmonary embolism (PE) who had SPECT-P and 4DCT imaging acquired within 24 h of PE diagnosis. The second cohort was comprised of 15 nonsmall cell lung cancer patients who had SPECT-P and 4DCT images acquired prior to radiotherapy. For each test case, CT-P images were computed for 30 different uncertainty parameter values, uniformly sampled from the range [0.01, 0.125], and the Spearman correlation between the SPECT-P and the resulting CT-P images were computed. RESULTS: The median correlations between CT-P and SPECT-P taken over all 30 test cases ranged between 0.49 and 0.57 across the parameter sweep. For the optimal tolerance τ = 0.0385, the CT-P and SPECT-P correlations across all 30 test cases ranged between 0.02 and 0.82. A one-sample sign test was applied separately to the PE and lung cancer cohorts. A low Spearmen correlation of 15% was set as the null median value and two-sided alternative was tested. The PE patients showed a median correlation of 0.57 (IQR = 0.305). One-sample sign test was statistically significant with 96.5 % confidence interval: 0.20-0.63, P < 0.00001. Lung cancer patients had a median correlation of 0.57(IQR = 0.230). Again, a one-sample sign test for median was statistically significant with 96.5 percent confidence interval: 0.45-0.71, P < 0.00001. CONCLUSION: CT-Perfusion is the first mechanistic model designed to quantify magnitude blood mass changes on noncontrast dynamic CT as a surrogate for pulmonary perfusion. While the reported correlations with SPECT-P are promising, further investigation is required to determine the optimal CT acquisition protocol and numerical method implementation for CT-P imaging.

Evaluating Positron Emission Tomography-Based Functional Imaging Changes in the Heart After Chemo-Radiation for Patients With Lung Cancer.

PURPOSE: Studies have noted a link between radiation dose to the heart and overall survival (OS) for patients with lung cancer treated with chemoradiation. The purpose of this study was to characterize pre- to posttreatment cardiac metabolic changes using fluorodeoxyglucose/positron emission tomography (FDG-PET) images and to evaluate whether changes in cardiac metabolism predict for OS. METHODS AND MATERIALS: Thirty-nine patients enrolled in a functional avoidance prospective study who had undergone pre- and postchemoradiation FDG-PET imaging were evaluated. For each patient, the pretreatment and posttreatment PET/CTs were rigidly registered to the planning CT, dose, and structure set. PET-based metabolic dose-response was assessed by comparing pretreatment to posttreatment mean standardized uptake values (SUVmean) in the heart as a function of dose-bin. OS analysis was performed by comparing SUVmean changes for patients who were alive or had died at last follow-up and by using a multivariate model to assess whether pre- to posttreatment SUVmean changes were a predictor of OS. RESULTS: The dose-response curve revealed increasing changes in SUV as a function of cardiac dose with an average SUVmean increase of 1.7% per 10 Gy. Patients were followed for a median of 437 days (range, 201-1131 days). SUVmean change was significantly predictive of OS on multivariate analysis with a hazard ratio of 0.541 (95% confidence intervals, 0.312-0.937). Patients alive at follow-up had an average increase of 17.2% in cardiac SUVmean while patients that died had an average decrease in SUVmean decrease of 13.5% (P = .048). CONCLUSIONS: Our data demonstrated that posttreatment SUV changes in the heart were significant indicators of dose-response and predictors of OS. The present work is hypothesis generating and must be validated in an independent cohort. If validated, our data show the potential for cardiac metabolic changes to be an early predictor for clinical outcomes.

Robust CT ventilation from the integral formulation of the Jacobian.

Computed tomography (CT) derived ventilation algorithms estimate the apparent voxel volume changes within an inhale/exhale CT image pair. Transformation-based methods compute these estimates solely from the spatial transformation acquired by applying a deformable image registration (DIR) algorithm to the image pair. However, approaches based on finite difference approximations of the transformation's Jacobian have been shown to be numerically unstable. As a result, transformation-based CT ventilation is poorly reproducible with respect to both DIR algorithm and CT acquisition method. PURPOSE: We introduce a novel Integrated Jacobian Formulation (IJF) method for estimating voxel volume changes under a DIR-recovered spatial transformation. The method is based on computing volume estimates of DIR mapped subregions using the hit-or-miss sampling algorithm for integral approximation. The novel approach allows for regional volume change estimates that (a) respect the resolution of the digital grid and (b) are based on approximations with quantitatively characterized and controllable levels of uncertainty. As such, the IJF method is designed to be robust to variations in DIR solutions and thus overall more reproducible. METHODS: Numerically, Jacobian estimates are recovered by solving a simple constrained linear least squares problem that guarantees the recovered global volume change is equal to the global volume change obtained from the inhale and exhale lung segmentation masks. Reproducibility of the IJF method with respect to DIR solution was assessed using the expert-determined landmark point pairs and inhale/exhale phases from 10 four-dimensional computed tomographies (4DCTs) available on www.dir-lab.com. Reproducibility with respect to CT acquisition was assessed on the 4DCT and 4D cone beam CT (4DCBCT) images acquired for five lung cancer patients prior to radiotherapy. RESULTS: The ten Dir-Lab 4DCT cases were registered twice with the same DIR algorithm, but with different smoothing parameter. Finite difference Jacobian (FDJ) and IFJ images were computed for both solutions. The average spatial errors (300 landmarks per case) for the two DIR solution methods were 0.98 (1.10) and 1.02 (1.11). The average Pearson correlation between the FDJ images computed from the two DIR solutions was 0.83 (0.03), while for the IJF images it was 1.00 (0.00). For intermodality assessment, the IJF and FDJ images were computed from the 4DCT and 4DCBCT of five patients. The average Pearson correlation of the spatially aligned FDJ images was 0.27 (0.11), while it was 0.77 (0.13) for the IFJ method. CONCLUSION: The mathematical theory underpinning the IJF method allows for the generation of ventilation images that are (a) computed with respect to DIR spatial accuracy on the digital voxel grid and (b) based on DIR-measured subregional volume change estimates acquired with quantifiable and controllable levels of uncertainty. Analyses of the experiments are consistent with the mathematical theory and indicate that IJF ventilation imaging has a higher reproducibility with respect to both DIR algorithm and CT acquisition method, in comparison to the standard finite difference approach.

Pulmonary blood mass dynamics on 4DCT during tidal breathing.

We previously reported that apparent lung mass varies across the phases of 4D computed tomography (4DCT) images. We hypothesize that these variations correspond to the physiologic changes in pulmonary perfusion induced during normal tidal breathing, and should therefore be present in every breathing patient. In this study, we characterize and quantify the respiratory induced variation in pulmonary blood mass (▵PBM) on 89 patients treated with stereotactic body radiotherapy. ▵PBM was computed from the treatment planning helical 4DCT images of each patient. Conversion from Hounsfield Units (HU) to density and mass per voxel was made using the density calibration curve, applied to the lung parenchyma volume within each phase. A difference in the lung mass with breathing was found for all cases, as was a substantial individual variation in lung volume. We found that the ▵PBM increased during inhalation, and decreased during exhalation. A significant correlation between the individual ▵PBM and tidal volume was observed; ▵PBM increased with tidal volume. We further evaluated the anatomic distribution of ▵PBM variation comparing the central versus peripheral lung, cranial versus caudal, dependent versus non-dependent lung. Our observations regarding spatial distribution of the ▵PBM agree with previously reported differences among similar regions for the supine patient. These results show that a variation in pulmonary mass during respiration is apparent on 4DCT and suggest that these variations reflect respiratory induced changes in the pulmonary perfusion. Therefore, the 4DCT derived respiratory induced ▵PBM signal can provide further insight into the pulmonary circulation and advance the overall understanding and diagnosis of human health and disease.