An adversarial machine learning framework and biomechanical model-guided approach for computing 3D lung tissue elasticity from end-expiration 3DCT.

PURPOSE: Lung elastography aims at measuring the lung parenchymal tissue elasticity for applications ranging from diagnostic purposes to biomechanically guided deformations. Characterizing the lung tissue elasticity requires four-dimensional (4D) lung motion as an input, which is currently estimated by deformably registering 4D computed tomography (4DCT) datasets. Since 4DCT imaging is widely used only in a radiotherapy treatment setup, there is a need to predict the elasticity distribution in the absence of 4D imaging for applications within and outside of radiotherapy domain. METHODS: In this paper, we present a machine learning-based method that predicts the three-dimensional (3D) lung tissue elasticity distribution for a given end-expiration 3DCT. The method to predict the lung tissue elasticity from an end-expiration 3DCT employed a deep neural network that predicts the tissue elasticity for the given CT dataset. For training and validation purposes, we employed five-dimensional CT (5DCT) datasets and a finite element biomechanical lung model. The 5DCT model was first used to generate end-expiration lung geometry, which was taken as the source lung geometry for biomechanical modeling. The deformation vector field pointing from end expiration to end inhalation was computed from the 5DCT model and taken as input in order to solve for the lung tissue elasticity. An inverse elasticity estimation process was employed, where we iteratively solved for the lung elasticity distribution until the model reproduced the ground-truth deformation vector field. The machine learning process uses a specific type of learning process, namely a constrained generalized adversarial neural network (cGAN) that learned the lung tissue elasticity in a supervised manner. The biomechanically estimated tissue elasticity together with the end-exhalation CT was the input for the supervised learning. The trained cGAN generated the elasticity from a given breath-hold CT image. The elasticity estimated was validated in two approaches. In the first approach, a L2-norm-based direct comparison was employed between the estimated elasticity and the ground-truth elasticity. In the second approach, we generated a synthetic four-dimensional CT (4DCT0 using a lung biomechanical model and the estimated elasticity and compared the deformations with the ground-truth 4D deformations using three image similarity metrics: mutual Information (MI), structured similarity index (SSIM), and normalized cross correlation (NCC). RESULTS: The results show that a cGAN-based machine learning approach was effective in computing the lung tissue elasticity given the end-expiration CT datasets. For the training data set, we obtained a learning accuracy of 0.44 ± 0.2 KPa. For the validation dataset, consisting of 13 4D datasets, we were able to obtain an accuracy of 0.87 ± 0.4 KPa. These results show that the cGAN-generated elasticity correlates well with that of the underlying ground-truth elasticity. We then integrated the estimated elasticity with the biomechanical model and applied the same boundary conditions in order to generate the end inhalation CT. The cGAN-generated images were very similar to that of the original end inhalation CT. The average value of the MI is 1.77 indicating the high local symmetricity between the ground truth and the cGAN elasticity-generated end inhalation CT data. The average value of the structural similarity for the 13 patients was observed to be 0.89 indicating the high structural integrity of the cGAN elasticity-generated end inhalation CT. Finally, the average NCC value of 0.97 indicates that potential variations in the contrast and brightness of the cGAN elasticity-generated end inhalation CT and the ground-truth end inhalation CT. CONCLUSION: The cGAN-generated lung tissue elasticity given an end-expiration CT image can be computed in near real time. Using the lung tissue elasticity along with a biomechanical model, 4D lung deformations can be generated from a given end-expiration CT image within clinically acceptable numerical accuracy.

A quantitative analysis of biomechanical lung model consistency using 5DCT datasets.

PURPOSE: Lung biomechanical models are important for understanding and characterizing lung anatomy and physiology. A key parameter of biomechanical modeling is the underlying tissue elasticity distribution. While human lung elasticity estimations do not have ground truths, model consistency checks can and should be employed to gauge the stability of the estimation techniques. This work proposes such a consistency check using a set of 10 subjects. METHODS: We hypothesize that lung dynamics will be stable over a 2-3 min time period and that this stability can be employed to check biomechanical estimation stability. For this purpose, two sets of 12 fast helical free breathing computed tomography scans (FHFBCT) were acquired back-to-back for each of the subjects. A published breathing motion model [five-dimensional CT (5DCT)] was generated from each set. Both of the models were used to generate two biomechanical modeling input sets: (a) The lung geometry at the end-exhalation, and (b) the voxel displacement map that mapped the end-exhalation lung geometry with the end-inhalation lung geometry. Finite element biomechanical lung models were instantiated using the end-exhalation lung geometries. The models included voxel-specific lung tissue elasticity values and were optimized using a gradient search approach until the biomechanical model-generated displacement maps matched those of the 5DCT voxel displacement maps. Finally, the two elasticity distributions associated with each of the patient 5DCTs were quantitatively compared. Because the end-exhalation geometries differed slightly between the two scan datasets, the elasticity distributions were deformably mapped to one of the exhalation datasets. RESULTS: For the 10 patients, on average, 90% of parenchymal voxels had <2 kPa Young's modulus difference between the two estimations, with a mean voxel difference of only 0.6 kPa. Similarly, 97% of the parenchymal voxels had <2 mm displacement difference between the two models with a mean difference of 0.48 mm. Furthermore, overlapping elasticity histograms for voxels between -600 and -900 HU (parenchymal tissues) showed that the histograms were consistent between the two estimations. CONCLUSION: In this paper, we demonstrated that biomechanical lung models can be consistently estimated when using motion-model based imaging datasets, even though the models were created from scans acquired at different breaths.

Systematic feasibility analysis of performing elastography using reduced dose CT lung image pairs.

PURPOSE: Elastography using computer tomography (CT) is a promising methodology that can provide patient-specific regional distributions of lung biomechanical properties. The purpose of this paper is to investigate the feasibility of performing elastography using simulated lower dose CT scans. METHODS: A cohort of eight patient CT image pairs were acquired with a tube current-time product of 40 mAs for estimating baseline lung elastography results. Synthetic low mAs CT scans were generated from the baseline scans to simulate the additional noise that would be present in acquisitions at 30, 25, and 20 mAs, respectively. For the simulated low mAs scans, exhalation and inhalation datasets were registered using an in-house optical flow deformable image registration algorithm. The registered deformation vector fields (DVFs) were taken to be ground truth for the elastography process. A model-based elasticity estimation was performed for each of the reduced mAs datasets, in which the goal was to optimize the elasticity distribution that best represented their respective DVFs. The estimated elasticity and the DVF distributions of the reduced mAs scans were then compared with the baseline elasticity results for quantitative accuracy purposes. RESULTS: The DVFs for the low mAs and baseline scans differed from each other by an average of 1.41 mm, which can be attributed to the noise added by the simulated reduction in mAs. However, the elastography results using the DVFs from the reduced mAs scans were similar from the baseline results, with an average elasticity difference of 0.65, 0.71, and 0.76 kPa, respectively. This illustrates that elastography can provide equivalent results using low-dose CT scans. CONCLUSIONS: Elastography can be performed equivalently using CT image pairs acquired with as low as 20 mAs. This expands the potential applications of CT-based elastography.

Safety-oriented design of in-house software for new techniques: A case study using a model-based 4DCT protocol.

PURPOSE: In-house software is commonly employed to implement new imaging and therapy techniques before commercial solutions are available. Risk analysis methods, as detailed in the TG-100 report of the American Association of Physicists in Medicine, provide a framework for quality management of processes but offer little guidance on software design. In this work, we examine a novel model-based four-dimensional computed tomography (4DCT) protocol using the TG-100 approach and describe two additional methods for promoting safety of the associated in-house software. METHODS: To implement a previously published model-based 4DCT protocol, in-house software was necessary for tasks such as synchronizing a respiratory signal to computed tomography images, deformable image registration (DIR), model parameter fitting, and interfacing with a treatment planning system. A process map was generated detailing the workflow. Failure modes and effects analysis (FMEA) was performed to identify critical steps and guide quality interventions. Software system safety was addressed through writing "use cases," narratives that characterize the behavior of the software, for all major operations to elicit safety requirements. Safety requirements were codified using the easy approach to requirements syntax (EARS) to ensure testability and eliminate ambiguity. RESULTS: Sixty-one failure modes were identified and assigned risk priority numbers using FMEA. Resultant quality management interventions include integration of a comprehensive reporting and logging system into the software, mandating daily and monthly equipment quality assurance procedures, and a checklist to be completed at image acquisition. Use cases and resulting safety requirements informed the design of needed in-house software as well as a suite of tests performed during the image generation process. CONCLUSIONS: TG-100 methods were used to construct a process-level quality management program for a 4DCT imaging protocol. Two supplemental tools from the field of requirements engineering facilitated elicitation and codification of safety requirements that informed the design and testing of in-house software necessary to implement the protocol. These general tools can be applied to promote safety when in-house software is needed to bring new techniques to the clinic.

Feasibility of deriving a novel imaging biomarker based on patient-specific lung elasticity for characterizing the degree of COPD in lung SBRT patients.

OBJECTIVE:: Lung tissue elasticity is an effective spatial representation for Chronic Obstructive Pulmonary Disease phenotypes and pathophysiology. We investigated a novel imaging biomarker based on the voxel-by-voxel distribution of lung tissue elasticity. Our approach combines imaging and biomechanical modeling to characterize tissue elasticity. METHODS:: We acquired 4DCT images for 13 lung cancer patients with known COPD diagnoses based on GOLD 2017 criteria. Deformation vector fields (DVFs) from the deformable registration of end-inhalation and end-exhalation breathing phases were taken to be the ground-truth. A linear elastic biomechanical model was assembled from end-exhalation datasets with a density-guided initial elasticity distribution. The elasticity estimation was formulated as an iterative process, where the elasticity was optimized based on its ability to reconstruct the ground-truth. An imaging biomarker (denoted YM1-3) derived from the optimized elasticity distribution, was compared with the current gold standard, RA950 using confusion matrix and area under the receiver operating characteristic (AUROC) curve analysis. RESULTS:: The estimated elasticity had 90 % accuracy when representing the ground-truth DVFs. The YM1-3 biomarker had higher diagnostic accuracy (86% vs 71 %), higher sensitivity (0.875 vs 0.5), and a higher AUROC curve (0.917 vs 0.875) as compared to RA950. Along with acting as an effective spatial indicator of lung pathophysiology, the YM1-3 biomarker also proved to be a better indicator for diagnostic purposes than RA950. CONCLUSIONS:: Overall, the results suggest that, as a biomarker, lung tissue elasticity will lead to new end points for clinical trials and new targeted treatment for COPD subgroups. ADVANCES IN KNOWLEDGE:: The derivation of elasticity information directly from 4DCT imaging data is a novel method for performing lung elastography. The work demonstrates the need for a mechanics-based biomarker for representing lung pathophysiology.

Estimation and validation of patient-specific high-resolution lung elasticity derived from 4DCT.

PURPOSE: Lung diseases are commonly associated with changes in lung tissue's biomechanical properties. Functional imaging techniques, such as elastography, have shown great promise in measuring tissue's biomechanical properties, which could expand the utility and effectiveness of radiotherapy treatment planning. We present a novel methodology for characterizing a key biomechanical property, parenchymal elasticity, derived solely from 4DCT datasets. METHODS: Specifically, end-inhalation and end-exhalation breathing phases of the 4DCT datasets were deformably registered and the resulting displacement maps were considered to be ground-truth. A mid-exhalation image was also prepared for verification purposes. A GPU-based biomechanical model was then generated from the patient end-exhalation dataset and used as a forward model to iteratively solve for the elasticity distribution. Displacements at the surface of the lungs were applied as boundary constraints for the model-guided tissue elastography, while the inner voxels were allowed to deform according to the linear elastic forces within the biomechanical model. A convergence criteria of 10% maximum deformation was employed for the iterative process. RESULTS: The lung tissue elasticity estimation was documented for a set of 15 4DCT patient datasets. Maximum lung deformations were observed to be between 6 and 31 mm. Our results showed that, on average, 89.91 ± 4.85% convergence was observed. A validation study consisting of mid-exhalation breathing phases illustrated an accuracy of 87.13 ± 10.62%. Structural similarity, normalized cross-correlation, and mutual information were used to quantify the degree of similarity between the following image pairs: (a) the model-generated end-exhalation and ground-truth end-exhalation, and (b) model-generated mid-exhalation images and ground-truth mid-exhalation. CONCLUSIONS: Overall, the results suggest that the lung elasticity can be measured with approximately 90% convergence using routinely acquired clinical 4DCT scans, indicating the potential for a lung elastography implementation within the radiotherapy clinical workflow. The regional lung elasticity found here can lead to improved tissue sparing radiotherapy treatment plans, and more precise monitoring of treatment response.

Systematic feasibility analysis of a quantitative elasticity estimation for breast anatomy using supine/prone patient postures.

PURPOSE: Breast elastography is a critical tool for improving the targeted radiotherapy treatment of breast tumors. Current breast radiotherapy imaging protocols only involve prone and supine CT scans. There is a lack of knowledge on the quantitative accuracy with which breast elasticity can be systematically measured using only prone and supine CT datasets. The purpose of this paper is to describe a quantitative elasticity estimation technique for breast anatomy using only these supine/prone patient postures. Using biomechanical, high-resolution breast geometry obtained from CT scans, a systematic assessment was performed in order to determine the feasibility of this methodology for clinically relevant elasticity distributions. METHODS: A model-guided inverse analysis approach is presented in this paper. A graphics processing unit (GPU)-based linear elastic biomechanical model was employed as a forward model for the inverse analysis with the breast geometry in a prone position. The elasticity estimation was performed using a gradient-based iterative optimization scheme and a fast-simulated annealing (FSA) algorithm. Numerical studies were conducted to systematically analyze the feasibility of elasticity estimation. For simulating gravity-induced breast deformation, the breast geometry was anchored at its base, resembling the chest-wall/breast tissue interface. Ground-truth elasticity distributions were assigned to the model, representing tumor presence within breast tissue. Model geometry resolution was varied to estimate its influence on convergence of the system. A priori information was approximated and utilized to record the effect on time and accuracy of convergence. The role of the FSA process was also recorded. A novel error metric that combined elasticity and displacement error was used to quantify the systematic feasibility study. For the authors' purposes, convergence was set to be obtained when each voxel of tissue was within 1 mm of ground-truth deformation. RESULTS: The authors' analyses showed that a ∼97% model convergence was systematically observed with no-a priori information. Varying the model geometry resolution showed no significant accuracy improvements. The GPU-based forward model enabled the inverse analysis to be completed within 10-70 min. Using a priori information about the underlying anatomy, the computation time decreased by as much as 50%, while accuracy improved from 96.81% to 98.26%. The use of FSA was observed to allow the iterative estimation methodology to converge more precisely. CONCLUSIONS: By utilizing a forward iterative approach to solve the inverse elasticity problem, this work indicates the feasibility and potential of the fast reconstruction of breast tissue elasticity using supine/prone patient postures.