Investigating the use of machine learning to generate ventilation images from CT scans.

BACKGROUND: Radiotherapy treatment planning incorporating ventilation imaging can reduce the incidence of radiation-induced lung injury. The gold-standard of ventilation imaging, using nuclear medicine, has limitations with respect to availability and cost. PURPOSE: An alternative type of ventilation imaging to nuclear medicine uses 4DCT (or breath-hold CT [BHCT] pair) with deformable image registration (DIR) and a ventilation metric to produce a CT ventilation image (CTVI). The purpose of this study is to investigate the application of machine learning as an alternative to DIR-based methods when producing CTVIs. METHODS: A patient dataset of 15 inhale and exhale BHCTs and Galligas PET ventilation images were used to train and test a 2D U-Net style convolutional neural network. The neural network established relationships between axial input BHCT image pairs and axial labeled Galligas PET images and was evaluated using eightfold cross-validation. Once trained, the neural network could produce a CTVI from an input BHCT image pair. The CTVIs produced by the neural network were qualitatively assessed visually and quantitatively compared to a Galligas PET ventilation image using a Spearman correlation and Dice similarity coefficient (DSC). The DSC measured the spatial overlap between three segmented equal lung volumes by ventilation (high, medium, and low functioning lung [LFL]). RESULTS: The mean Spearman correlation between the CTVIs and the Galligas PET ventilation images was 0.58 ± 0.14. The mean DSC over high, medium, and LFL between the CTVIs and Galligas PET ventilation images was 0.55 ± 0.06. Visually, a systematic overprediction of ventilation within the lung was observed in the CTVIs with respect to the Galligas PET ventilation images, with jagged regions of ventilation in the sagittal and coronal planes. CONCLUSIONS: A convolutional neural network was developed that could produce a CTVI from a BHCT image pair, which was then compared with a Galligas PET ventilation image. The performance of this machine learning method was comparable to previous benchmark studies investigating a DIR-based CTVI, warranting future development, and investigation of applying machine learning to a CTVI.

Combining Serial and Parallel Functionality in Functional Lung Avoidance Radiation Therapy.

PURPOSE: Functional lung avoidance (FLA) radiation therapy (RT) aims to minimize post-RT pulmonary toxicity by preferentially avoiding dose to high-functioning lung (HFL) regions. A common limitation is that FLA approaches do not consider the conducting architecture for gas exchange. We previously proposed the functionally weighted airway sparing (FWAS) method to spare airways connected to HFL regions, showing that it is possible to substantially reduce risk of radiation-induced airway injury. Here, we compare the performance of FLA and FWAS and propose a novel method combining both approaches. METHODS: We used breath-hold computed tomography (BHCT) and simulation 4-dimensional computed tomography (4DCT) from 12 lung stereotactic ablative radiation therapy patients. Four planning strategies were examined: (1) Conventional: no sparing other than clinical dose-volume constraints; (2) FLA: using a 4DCT-based ventilation map to delineate the HFL, plans were optimized to reduce mean dose and V13.50 in HFL; (3) FWAS: we autosegemented 11 to 13 generations of individual airways from each patient's BHCT and assigned priorities based on the relative contribution of each airway to total ventilation. We used these priorities in the optimization along with airway dose constraints, estimated as a function of airway diameter and 5% probability of collapse; and (4) FLA + FWAS: we combined information from the 2 strategies. We prioritized clinical dose constraints for organs at risk and planning target volume in all plans. We performed the evaluation in terms of ventilation preservation accounting for radiation-induced damage to both lung parenchyma and airways. RESULTS: We observed average ventilation preservation for FLA, FWAS, and FLA + FWAS as 3%, 8.5%, and 14.5% higher, respectively, than for Conventional plans for patients with ventilation preservation in Conventional plans <90%. Generalized estimated equations showed that all improvements were statistically significant (P ≤ .036). We observed no clinically relevant improvements in outcomes of the sparing techniques in patients with ventilation preservation in Conventional plans ≥90%. CONCLUSIONS: These initial results suggest that it is crucial to consider the parallel and the serial nature of the lung to improve post-radiation therapy lung function and, consequently, quality of life for patients.

First experimental evaluation of multi-target multileaf collimator tracking during volumetric modulated arc therapy for locally advanced prostate cancer.

PURPOSE: Locally advanced and oligometastatic cancer patients require radiotherapy treatment to multiple independently moving targets. There is no existing commercial solution that can simultaneously track and treat multiple targets. This study experimentally implemented and evaluated a real-time multi-target tracking system for locally advanced prostate cancer. METHODS: Real-time multi-target MLC tracking was integrated with 3D x-ray image guidance on a standard linac. Three locally advanced prostate cancer treatment plans were delivered to a static lymph node phantom and dynamic prostate phantom that reproduced three prostate trajectories. Treatments were delivered using multi-target MLC tracking, single-target MLC tracking, and no tracking. Doses were measured using Gafchromic film placed in the dynamic and static phantoms. Dosimetric error was quantified by the 2%/2 mm gamma failure rate. Geometric error was evaluated as the misalignment between target and aperture positions. The multi-target tracking system latency was measured. RESULTS: The mean (range) gamma failure rates for the prostate and lymph nodes, were 18.6% (5.2%, 28.5%) and 7.5% (1.1%, 13.7%) with multi-target tracking, 7.9% (0.7%, 15.4%) and 37.8% (18.0%, 57.9%) with single-target tracking, and 38.1% (0.6%, 75.3%) and 37.2% (29%, 45.3%) without tracking. Multi-target tracking had the lowest geometric error with means and standard deviations within 0.2 ± 1.5 for the prostate and 0.0 ± 0.3 mm for the lymph nodes. The latency was 730 ± 20 ms. CONCLUSION: This study presented the first experimental implementation of multi-target tracking to independently track prostate and lymph node displacement during VMAT. Multi-target tracking reduced dosimetric and geometric errors compared to single-target tracking and no tracking.

Rolling out RapidPlan: What we've learnt.

INTRODUCTION: RapidPlan (RP), a knowledge-based planning system, aims to consistently improve plan quality and efficiency in radiotherapy. During the early stages of implementation, some of the challenges include knowing how to optimally train a model and how to integrate RP into a department. We discuss our experience with the implementation of RP into our institution. METHODS: We reviewed all patients planned using RP over a 7-month period following inception in our department. Our primary outcome was clinically acceptable plans (used for treatment) with secondary outcomes including model performance and a comparison of efficiency and plan quality between RP and manual planning (MP). RESULTS: Between November 2017 and May 2018, 496 patients were simulated, of which 217 (43.8%) had an available model. RP successfully created a clinically acceptable plan in 87.2% of eligible patients. The individual success of the 24 models ranged from 50% to 100%, with more than 90% success in 15 (62.5%) of the models. In 40% of plans, success was achieved on the 1st optimisation. The overall planning time with RP was reduced by up to 95% compared with MP times. The quality of the RP plans was at least equivalent to historical MP plans in terms of target coverage and organ at risk constraints. CONCLUSION: While initially time-consuming and resource-intensive to implement, plans optimised with RP demonstrate clinically acceptable plan quality, while significantly improving the efficiency of a department, suggesting RP and its application is a highly effective tool in clinical practice.

Real-Time Image Guided Ablative Prostate Cancer Radiation Therapy: Results From the TROG 15.01 SPARK Trial.

PURPOSE: Kilovoltage intrafraction monitoring (KIM) is a novel software platform implemented on standard radiation therapy systems and enabling real-time image guided radiation therapy (IGRT). In a multi-institutional prospective trial, we investigated whether real-time IGRT improved the accuracy of the dose patients with prostate cancer received during radiation therapy. METHODS AND MATERIALS: Forty-eight patients with prostate cancer were treated with KIM-guided SABR with 36.25 Gy in 5 fractions. During KIM-guided treatment, the prostate motion was corrected for by either beam gating with couch shifts or multileaf collimator tracking. A dose reconstruction method was used to evaluate the dose delivered to the target and organs at risk with and without real-time IGRT. Primary outcome was the effect of real-time IGRT on dose distributions. Secondary outcomes included patient-reported outcomes and toxicity. RESULTS: Motion correction occurred in ≥1 treatment for 88% of patients (42 of 48) and 51% of treatments (121 of 235). With real-time IGRT, no treatments had prostate clinical target volume (CTV) D98% dose 5% less than planned. Without real-time IGRT, 13 treatments (5.5%) had prostate CTV D98% doses 5% less than planned. The prostate CTV D98% dose with real-time IGRT was closer to the plan by an average of 1.0% (range, -2.8% to 20.3%). Patient outcomes showed no change in the 12-month patient-reported outcomes compared with baseline and no grade ≥3 genitourinary or gastrointestinal toxicities. CONCLUSIONS: Real-time IGRT is clinically effective for prostate cancer SABR.

Imaging of regional ventilation: Is CT ventilation imaging the answer? A systematic review of the validation data.

Computed Tomography Ventilation Imaging (CTVI) is an experimental imaging modality that derives regional lung function information from non-contrast respiratory-correlated CT datasets. Despite CTVI being extensively studied in cross-modality imaging comparisons, there is a lack of consensus on the state of its clinical validation in humans. This systematic review evaluates the CTVI clinical validation studies to date, highlights their common strengths and weaknesses and makes recommendations. We performed a PUBMED and EMBASE search of all English language papers on CTVI between 2000 and 2018. The results of these searches were filtered in accordance to a set of eligibility criteria and analysed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Guidelines. One hundred and forty-four records were identified, and 66 full text records were reviewed. After detailed assessment, twenty-three full text papers met the selection criteria and were included in the final review. This included thirteen prospective studies, with 579 human subjects. Studies used diverse methodologies, with a large amount of heterogeneity between different studies in terms of the reference ventilation imaging modality (e.g. nuclear medicine, hyperpolarised gas MRI), imaging parameters, DIR algorithm(s) used, and ventilation metric(s) applied. The most common ventilation metrics used deformable image registration to evaluate the exhale-to-inhale motion field Jacobian determinant (DIR-Jac) or changes in air volume content based on Hounsfield Units (DIR-HU). The strength of correlation between CTVI and the reference ventilation imaging modalities was moderate to strong when evaluated at the lobar or global level, with the average ±â€¯S.D. (number of studies) linear regression correlation coefficients were 0.73 ±â€¯0.25 (n = 6) and 0.86 ±â€¯0.11 (n = 12) for DIR-Jac and DIR-HU respectively, and the SPC were 0.45 ±â€¯0.31 (n = 6) and 0.41 ±â€¯0.11 (n = 5) for DIR-Jac and DIR-HU respectively. We concluded that it is difficult to make a broad statement about the validity of CTVI due to the diverse methods used in the validation literature. Typically, CTVI appears to show reasonable cross-modality correlations at the lobar/whole lung level but poor correlations at the voxel level. Since CTVI is seeing new implementations in prospective trials, it is clear that refinement and standardization of the clinical validation methodologies are required. CTVI appears to be of relevance in radiotherapy planning, particularly in patients whose main pulmonary impairment is not a gas exchange problem but alternative imaging approaches may need to be considered in patients with other pulmonary diseases (i.e. restrictive or gas exchange problems).

Big Data Readiness in Radiation Oncology: An Efficient Approach for Relabeling Radiation Therapy Structures With Their TG-263 Standard Name in Real-World Data Sets.

PURPOSE: To prepare for big data analyses on radiation therapy data, we developed Stature, a tool-supported approach for standardization of structure names in existing radiation therapy plans. We applied the widely endorsed nomenclature standard TG-263 as the mapping target and quantified the structure name inconsistency in 2 real-world data sets. METHODS AND MATERIALS: The clinically relevant structures in the radiation therapy plans were identified by reference to randomized controlled trials. The Stature approach was used by clinicians to identify the synonyms for each relevant structure, which was then mapped to the corresponding TG-263 name. We applied Stature to standardize the structure names for 654 patients with prostate cancer (PCa) and 224 patients with head and neck squamous cell carcinoma (HNSCC) who received curative radiation therapy at our institution between 2007 and 2017. The accuracy of the Stature process was manually validated in a random sample from each cohort. For the HNSCC cohort we measured the resource requirements for Stature, and for the PCa cohort we demonstrated its impact on an example clinical analytics scenario. RESULTS: All but 1 synonym group ("Hydrogel") was mapped to the corresponding TG-263 name, resulting in a TG-263 relabel rate of 99% (8837 of 8925 structures). For the PCa cohort, Stature matched a total of 5969 structures. Of these, 5682 structures were exact matches (ie, following local naming convention), 284 were matched via a synonym, and 3 required manual matching. This original radiation therapy structure names therefore had a naming inconsistency rate of 4.81%. For the HNSCC cohort, Stature mapped a total of 2956 structures (2638 exact, 304 synonym, 14 manual; 10.76% inconsistency rate) and required 7.5 clinician hours. The clinician hours required were one-fifth of those that would be required for manual relabeling. The accuracy of Stature was 99.97% (PCa) and 99.61% (HNSCC). CONCLUSIONS: The Stature approach was highly accurate and had significant resource efficiencies compared with manual curation.

The VAMPIRE challenge: A multi-institutional validation study of CT ventilation imaging.

PURPOSE: CT ventilation imaging (CTVI) is being used to achieve functional avoidance lung cancer radiation therapy in three clinical trials (NCT02528942, NCT02308709, NCT02843568). To address the need for common CTVI validation tools, we have built the Ventilation And Medical Pulmonary Image Registration Evaluation (VAMPIRE) Dataset, and present the results of the first VAMPIRE Challenge to compare relative ventilation distributions between different CTVI algorithms and other established ventilation imaging modalities. METHODS: The VAMPIRE Dataset includes 50 pairs of 4DCT scans and corresponding clinical or experimental ventilation scans, referred to as reference ventilation images (RefVIs). The dataset includes 25 humans imaged with Galligas 4DPET/CT, 21 humans imaged with DTPA-SPECT, and 4 sheep imaged with Xenon-CT. For the VAMPIRE Challenge, 16 subjects were allocated to a training group (with RefVI provided) and 34 subjects were allocated to a validation group (with RefVI blinded). Seven research groups downloaded the Challenge dataset and uploaded CTVIs based on deformable image registration (DIR) between the 4DCT inhale/exhale phases. Participants used DIR methods broadly classified into B-splines, Free-form, Diffeomorphisms, or Biomechanical modeling, with CT ventilation metrics based on the DIR evaluation of volume change, Hounsfield Unit change, or various hybrid approaches. All CTVIs were evaluated against the corresponding RefVI using the voxel-wise Spearman coefficient rS , and Dice similarity coefficients evaluated for low function lung ( DSClow ) and high function lung ( DSChigh ). RESULTS: A total of 37 unique combinations of DIR method and CT ventilation metric were either submitted by participants directly or derived from participant-submitted DIR motion fields using the in-house software, VESPIR. The rS and DSC results reveal a high degree of inter-algorithm and intersubject variability among the validation subjects, with algorithm rankings changing by up to ten positions depending on the choice of evaluation metric. The algorithm with the highest overall cross-modality correlations used a biomechanical model-based DIR with a hybrid ventilation metric, achieving a median (range) of 0.49 (0.27-0.73) for rS , 0.52 (0.36-0.67) for DSClow , and 0.45 (0.28-0.62) for DSChigh . All other algorithms exhibited at least one negative rS value, and/or one DSC value less than 0.5. CONCLUSIONS: The VAMPIRE Challenge results demonstrate that the cross-modality correlation between CTVIs and the RefVIs varies not only with the choice of CTVI algorithm but also with the choice of RefVI modality, imaging subject, and the evaluation metric used to compare relative ventilation distributions. This variability may arise from the fact that each of the different CTVI algorithms and RefVI modalities provides a distinct physiologic measurement. Ultimately this variability, coupled with the lack of a "gold standard," highlights the ongoing importance of further validation studies before CTVI can be widely translated from academic centers to the clinic. It is hoped that the information gleaned from the VAMPIRE Challenge can help inform future validation efforts.

Image-based retrospective 4D MRI in external beam radiotherapy: A comparative study with a digital phantom.

PURPOSE: Several image-based retrospective sorting methods of 4D magnetic resonance imaging (4D MRI) have been proposed for respiratory motion reconstruction in external beam radiotherapy. However, the optimal strategy for providing accurate and artifact-free 4D MRI, ideally corresponding to an average breathing cycle, is not yet defined. This study presents a proactive comparison of three published image-based sorting methods, to define a groundwork for benchmarking in 4D MRI. METHODS: Three published 4D MRI methods were selected for image retrospective sorting: body area, mutual information, and navigator slice. The three image-based methods were compared against a conventional retrospective sorting method based on an external surrogate. Comparisons were performed by means of an MRI digital phantom, derived from the XCAT CT phantom generated with different patient-derived signals, for a total of 12 cases. Specific multislice MRI acquisitions were simulated for slice sorting and sagittal, coronal, and axial orientations were tested. An average 4D cycle was generated as ground truth. RESULTS: Individual and grouped patient analyses showed better performance of the navigator slice and mutual information in amplitude binning with respect to the body area strategy. Binning artifacts were reduced on the diaphragm with the slice navigator method due to the acquired internal information. Tumor motion description accurately matched the ground truth in the mutual information strategy with amplitude binning. The body area method followed the performance of the external surrogate and presented larger errors, since was not correlated with the internal anatomy. Sagittal and coronal orientations reported lower errors than axial slicing. Individual analysis showed the need of a patient-specific evaluation for the selection of the best method. CONCLUSIONS: A comparison between three different image-based retrospective sorting methods for 4D MRI is proposed, providing guidelines for benchmark definition in MRI-guided radiotherapy.

Feasibility study on 3D image reconstruction from 2D orthogonal cine-MRI for MRI-guided radiotherapy.

INTRODUCTION: In-room MRI is a promising image guidance strategy in external beam radiotherapy to acquire volumetric information for moving targets. However, limitations in spatio-temporal resolution led several authors to use 2D orthogonal images for guidance. The aim of this work is to present a method to concurrently compensate for non-rigid tumour motion and provide an approach for 3D reconstruction from 2D orthogonal cine-MRI slices for MRI-guided treatments. METHODS: Free-breathing sagittal/coronal interleaved 2D cine-MRI were acquired in addition to a pre-treatment 3D volume in two patients. We performed deformable image registration (DIR) between cine-MRI slices and corresponding slices in the pre-treatment 3D volume. Based on an extrapolation of the interleaved 2D motion fields, the 3D motion field was estimated and used to warp the pre-treatment volume. Due to the lack of a ground truth for patients, the method was validated on a digital 4D lung phantom. RESULTS: On the phantom, the 3D reconstruction method was able to compensate for tumour motion and compared favourably to the results of previously adopted strategies. The difference in the 3D motion fields between the phantom and the extrapolated motion was 0.4 ± 0.3 mm for tumour and 0.8 ± 1.5 mm for whole anatomy, demonstrating feasibility of performing a 3D volumetric reconstruction directly from 2D orthogonal cine-MRI slices. Application of the method to patient data confirmed the feasibility of utilizing this method in real world scenarios. CONCLUSION: Preliminary results on phantom and patient cases confirm the feasibility of the proposed approach in an MRI-guided scenario, especially for non-rigid tumour motion compensation.

CT ventilation imaging derived from breath hold CT exhibits good regional accuracy with Galligas PET.

BACKGROUND AND PURPOSE: CT ventilation imaging (CTVI) derived from four dimensional CT (4DCT) has shown only moderate spatial accuracy in humans due to 4DCT image artefacts. Here we assess the accuracy of an improved CTVI using high quality exhale/inhale breath-hold CT (BHCT). MATERIALS AND METHODS: Eighteen lung cancer patients underwent exhale/inhale BHCT, 4DCT and Galligas PET ventilation scans in a single imaging session. For each BHCT and 4DCT scan, we performed deformable image registration (DIR) between the inhale and exhale phase images to quantify ventilation using three published metrics: (i) breathing induced lung density change, CTVIDIR-HU (ii) breathing induced volume change CTVIDIR-Jac and (iii) the regional air-tissue product, CTVIHU Spatial accuracy was reported as the voxel-wise Spearman correlation r between CTVI and Galligas PET. RESULTS: For BHCT-based CTVIs (N = 16), the CTVIDIR-HU, CTVIDIR-Jac and CTVIHU methods yielded mean (range) r values of 0.67 (0.52-0.87), 0.57 (0.18-0.77) and 0.49 (0.14-0.75) respectively. By comparison the 4DCT-based CTVIs (n = 14) had values of 0.32 (-0.04 to 0.51), 0.16 (-0.31 to 44) and 0.49 (0.20-0.77) respectively. CONCLUSIONS: High quality CT imaging is a key requirement for accurate CT ventilation imaging. The use of exhale/inhale BHCT can improve the accuracy of CTVI for human subjects.

Evaluating the accuracy of 4D-CT ventilation imaging: First comparison with Technegas SPECT ventilation.

PURPOSE: Computed tomography ventilation imaging (CTVI) is a highly accessible functional lung imaging modality that can unlock the potential for functional avoidance in lung cancer radiation therapy. Previous attempts to validate CTVI against clinical ventilation single-photon emission computed tomography (V-SPECT) have been hindered by radioaerosol clumping artifacts. This work builds on those studies by performing the first comparison of CTVI with 99m Tc-carbon ('Technegas'), a clinical V-SPECT modality featuring smaller radioaerosol particles with less clumping. METHODS: Eleven lung cancer radiotherapy patients with early stage (T1/T2N0) disease received treatment planning four-dimensional CT (4DCT) scans paired with Technegas V/Q-SPECT/CT. For each patient, we applied three different CTVI methods. Two of these used deformable image registration (DIR) to quantify breathing-induced lung density changes (CTVIDIR-HU ), or breathing-induced lung volume changes (CTVIDIR-Jac ) between the 4DCT exhale/inhale phases. A third method calculated the regional product of air-tissue densities (CTVIHU ) and did not involve DIR. Corresponding CTVI and V-SPECT scans were compared using the Dice similarity coefficient (DSC) for functional defect and nondefect regions, as well as the Spearman's correlation r computed over the whole lung. The DIR target registration error (TRE) was quantified using both manual and computer-selected anatomic landmarks. RESULTS: Interestingly, the overall best performing method (CTVIHU ) did not involve DIR. For nondefect regions, the CTVIHU , CTVIDIR-HU , and CTVIDIR-Jac methods achieved mean DSC values of 0.69, 0.68, and 0.54, respectively. For defect regions, the respective DSC values were moderate: 0.39, 0.33, and 0.44. The Spearman r-values were generally weak: 0.26 for CTVIHU , 0.18 for CTVIDIR-HU , and -0.02 for CTVIDIR-Jac . The spatial accuracy of CTVI was not significantly correlated with TRE, however the DIR accuracy itself was poor with TRE > 3.6 mm on average, potentially indicative of poor quality 4DCT. Q-SPECT scans achieved good correlations with V-SPECT (mean r > 0.6), suggesting that the image quality of Technegas V-SPECT was not a limiting factor in this study. CONCLUSIONS: We performed a validation of CTVI using clinically available 4DCT and Technegas V/Q-SPECT for 11 lung cancer patients. The results reinforce earlier findings that the spatial accuracy of CTVI exhibits significant interpatient and intermethod variability. We propose that the most likely factor affecting CTVI accuracy was poor image quality of clinical 4DCT.

Quantifying the reproducibility of lung ventilation images between 4-Dimensional Cone Beam CT and 4-Dimensional CT.

PURPOSE: Computed tomography ventilation imaging derived from four-dimensional cone beam CT (CTVI4DCBCT ) can complement existing 4DCT-based methods (CTVI4DCT ) to track lung function changes over a course of lung cancer radiation therapy. However, the accuracy of CTVI4DCBCT needs to be assessed since anatomic 4DCBCT has demonstrably poor image quality and small field of view (FOV) compared to treatment planning 4DCT. We perform a direct comparison between short interval CTVI4DCBCT and CTVI4DCT pairs to understand the patient specific image quality factors affecting the intermodality CTVI reproducibility in the clinic. METHODS AND MATERIALS: We analysed 51 pairs of 4DCBCT and 4DCT scans acquired within 1 day of each other for nine lung cancer patients. To assess the impact of image quality, CTVIs were derived from 4DCBCT scans reconstructed using both standard Feldkamp-Davis-Kress backprojection (CTVIFDK4DCBCT) and an iterative McKinnon-Bates Simultaneous Algebraic Reconstruction Technique (CTVIMKBSART4DCBCT). Also, the influence of FOV was assessed by deriving CTVIs from 4DCT scans that were cropped to a similar FOV as the 4DCBCT scans (CTVIcrop4DCT), or uncropped (CTVIuncrop4DCT). All CTVIs were derived by performing deformable image registration (DIR) between the exhale and inhale phases and evaluating the Jacobian determinant of deformation. Reproducibility between corresponding CTVI4DCBCT and CTVI4DCT pairs was quantified using the voxel-wise Spearman rank correlation and the Dice similarity coefficient (DSC) for ventilation defect regions (identified as the lower quartile of ventilation values). Mann-Whitney U-tests were applied to determine statistical significance of each reconstruction and cropping condition. RESULTS: The (mean ± SD) Spearman correlation between CTVIFDK4DCBCT and CTVIuncrop4DCT was 0.60 ± 0.23 (range -0.03-0.88) and the DSC was 0.64 ± 0.12 (0.34-0.83). By comparison, correlations between CTVIMKBSART4DCBCT and CTVIuncrop4DCT showed a small but statistically significant improvement with = 0.64 ± 0.20 (range 0.06-0.90, P = 0.03) and DSC = 0.66 ± 0.13 (0.31-0.87, P = 0.02). Intermodal correlations were noted to decrease with an increasing fraction of lung truncation in 4DCBCT relative to 4DCT, albeit not significantly (Pearson correlation R = 0.58, P = 0.002). CONCLUSIONS: This study demonstrates that DIR based CTVIs derived from 4DCBCT can exhibit reasonable to good voxel-level agreement with CTVIs derived from 4DCT. These correlations outperform previous cross-modality comparisons between 4DCT-based ventilation and nuclear medicine. The use of 4DCBCT scans with iterative reconstruction and minimal lung truncation is recommended to ensure better reproducibility between 4DCBCT- and 4DCT-based CTVIs.

The impact of breathing guidance and prospective gating during thoracic 4DCT imaging: an XCAT study utilizing lung cancer patient motion.

Two interventions to overcome the deleterious impact irregular breathing has on thoracic-abdominal 4D computed tomography (4DCT) are (1) facilitating regular breathing using audiovisual biofeedback (AVB), and (2) prospective respiratory gating of the 4DCT scan based on the real-time respiratory motion. The purpose of this study was to compare the impact of AVB and gating on 4DCT imaging using the 4D eXtended cardiac torso (XCAT) phantom driven by patient breathing patterns. We obtained simultaneous measurements of chest and abdominal walls, thoracic diaphragm, and tumor motion from 6 lung cancer patients under two breathing conditions: (1) AVB, and (2) free breathing. The XCAT phantom was used to simulate 4DCT acquisitions in cine and respiratory gated modes. 4DCT image quality was quantified by artefact detection (NCCdiff), mean square error (MSE), and Dice similarity coefficient of lung and tumor volumes (DSClung, DSCtumor). 4DCT acquisition times and imaging dose were recorded. In cine mode, AVB improved NCCdiff, MSE, DSClung, and DSCtumor by 20% (p = 0.008), 23% (p < 0.001), 0.5% (p < 0.001), and 4.0% (p < 0.003), respectively. In respiratory gated mode, AVB improved NCCdiff, MSE, and DSClung by 29% (p < 0.001), 34% (p < 0.001), 0.4% (p < 0.001), respectively. AVB increased the cine acquisitions by 15 s and reduced respiratory gated acquisitions by 31 s. AVB increased imaging dose in cine mode by 10%. This was the first study to quantify the impact of breathing guidance and respiratory gating on 4DCT imaging. With the exception of DSCtumor in respiratory gated mode, AVB significantly improved 4DCT image analysis metrics in both cine and respiratory gated modes over free breathing. The results demonstrate that AVB and respiratory-gating can be beneficial interventions to improve 4DCT for cancer radiation therapy, with the biggest gains achieved when these interventions are used simultaneously.

Estimating lung ventilation directly from 4D CT Hounsfield unit values.

PURPOSE: Computed tomography ventilation imaging (CTVI) aims to visualize air-volume changes in the lung by quantifying respiratory motion in 4DCT using deformable image registration (DIR). A problem is that DIR-based CTVI is sensitive both to 4DCT image artifacts and DIR parameters, hindering clinical validation of the technique. To address this, the authors present a streamlined CTVI approach that estimates blood-gas exchange in terms of time-averaged 4DCT Hounsfield unit (HU) values without relying on DIR. The purpose of this study is to quantify the accuracy of the HU-based CTVI method using high-resolution (68)Ga positron emission tomography ("Galligas PET") scans in lung cancer patients. METHODS: The authors analyzed Galligas 4D-PET/CT scans acquired for 25 lung cancer patients at up to three imaging timepoints during lung cancer radiation therapy. For each 4DCT scan, the authors produced three types of CTVIs: (i) the new method (CTV IHU¯), which takes the 4D time-averaged product of regional air and tissue densities at each voxel, and compared this to DIR-based estimates of (ii) breathing-induced density changes (CTV IDIR-HU), and (iii) breathing-induced volume changes (CTV IDIR-Jac) between the exhale/inhale phase images. The authors quantified the accuracy of CTV IHU¯, CTV IDIR-HU and CTV IDIR-Jac versus Galligas PET in terms of voxel-wise Spearman correlation (r) and the separation of mean voxel values between clinically defined defect/nondefect regions. RESULTS: Averaged over 62 scans, CTV IHU¯ showed better accuracy than CTV IDIR-HU and CTV IDIR-Jac in terms of Spearman correlation with Galligas PET, with (mean ± SD) r values of (0.50 ± 0.17), (0.42 ± 0.20), and (0.19 ± 0.23), respectively. A two-sample Kolmogorov-Smirnov test indicates that CTV IHU¯ shows statistically significant separation of mean ventilation values between clinical defect/nondefect regions. Qualitatively, CTV IHU¯ appears concordant with Galligas PET for emphysema related defects, but differences arise in tumor-obstructed regions (where aeration is overestimated due to motion blur) and for other abnormal morphology (e.g., fluid-filled or peritumoral lung with HU ≳ - 600) where the assumptions of the HU model may break down. CONCLUSIONS: The HU-based CTVI method can improve voxel-wise correlations with Galligas PET compared to DIR-based methods and may be a useful approximation for voxels with HU values in the range (-1000, - 600). With further clinical verification, HU-based CTVI could provide a straightforward and reproducible means to estimate lung ventilation using free-breathing 4DCT.

Measurement of preoperative lobar lung function with computed tomography ventilation imaging: progress towards rapid stratification of lung cancer lobectomy patients with abnormal lung function.

OBJECTIVES: In lung cancer preoperative evaluation, functional lung imaging is commonly used to assess lobar function. Computed tomography ventilation (CT-V) imaging is an emerging lung function imaging modality. We compared CT-V imaging assessment of lobar function and its prediction of postoperative lung function to that achieved by (i) positron emission tomography ventilation (PET-V) imaging and (ii) the standard anatomical segment counting (ASC) method. We hypothesized (i) that CT-V and PET-V have similar relative lobar function and (ii) that functional imaging and anatomic assessment (ASC) yield different predicted postoperative (ppo) lung function and therefore could change clinical management. METHODS: In this proof-of-concept study, 11 patients were subjected to pulmonary function tests, CT-V and PET-V imaging. The Bland-Altman plot, Pearson's correlation and linear regression analysis were used to assess the agreement between the CT-V-, PET-V- and ASC-based quantification of lobar function and in the ppo lung function. RESULTS: CT-V and PET-V imaging demonstrated strong correlations in quantifying relative lobar function (r = 0.96; P < 0.001). A Wilcoxon-signed rank test showed no significant difference in the lobar function estimates between the two imaging modalities (P = 0.83). The Bland-Altman plot also showed no significant differences. The correlation between ASC-based lobar function estimates with ventilation imaging was low, r < 0.45; however, the predictions of postoperative lung function correlated strongly between all three methods. CONCLUSIONS: The assessment of lobar function from CT-V imaging correlated strongly with PET-V imaging, but had low correlations with ASC. CT-V imaging may be a useful alternative method in preoperative evaluation for lung cancer patients.

Quantifying the image quality and dose reduction of respiratory triggered 4D cone-beam computed tomography with patient-measured breathing.

Respiratory triggered four dimensional cone-beam computed tomography (RT 4D CBCT) is a novel technique that uses a patient's respiratory signal to drive the image acquisition with the goal of imaging dose reduction without degrading image quality. This work investigates image quality and dose using patient-measured respiratory signals for RT 4D CBCT simulations. Studies were performed that simulate a 4D CBCT image acquisition using both the novel RT 4D CBCT technique and a conventional 4D CBCT technique. A set containing 111 free breathing lung cancer patient respiratory signal files was used to create 111 pairs of RT 4D CBCT and conventional 4D CBCT image sets from realistic simulations of a 4D CBCT system using a Rando phantom and the digital phantom, XCAT. Each of these image sets were compared to a ground truth dataset from which a mean absolute pixel difference (MAPD) metric was calculated to quantify the degradation of image quality. The number of projections used in each simulation was counted and was assumed as a surrogate for imaging dose. Based on 111 breathing traces, when comparing RT 4D CBCT with conventional 4D CBCT, the average image quality was reduced by 7.6% (Rando study) and 11.1% (XCAT study). However, the average imaging dose reduction was 53% based on needing fewer projections (617 on average) than conventional 4D CBCT (1320 projections). The simulation studies have demonstrated that the RT 4D CBCT method can potentially offer a 53% saving in imaging dose on average compared to conventional 4D CBCT in simulation studies using a wide range of patient-measured breathing traces with a minimal impact on image quality.

Measuring interfraction and intrafraction lung function changes during radiation therapy using four-dimensional cone beam CT ventilation imaging.

PURPOSE: Adaptive ventilation guided radiation therapy could minimize the irradiation of healthy lung based on repeat lung ventilation imaging (VI) during treatment. However the efficacy of adaptive ventilation guidance requires that interfraction (e.g., week-to-week), ventilation changes are not washed out by intrafraction (e.g., pre- and postfraction) changes, for example, due to patient breathing variability. The authors hypothesize that patients undergoing lung cancer radiation therapy exhibit larger interfraction ventilation changes compared to intrafraction function changes. To test this, the authors perform the first comparison of interfraction and intrafraction lung VI pairs using four-dimensional cone beam CT ventilation imaging (4D-CBCT VI), a novel technique for functional lung imaging. METHODS: The authors analyzed a total of 215 4D-CBCT scans acquired for 19 locally advanced non-small cell lung cancer (LA-NSCLC) patients over 4-6 weeks of radiation therapy. This set of 215 scans was sorted into 56 interfraction pairs (including first day scans and each of treatment weeks 2, 4, and 6) and 78 intrafraction pairs (including pre/postfraction scans on the same-day), with some scans appearing in both sets. VIs were obtained from the Jacobian determinant of the transform between the 4D-CBCT end-exhale and end-inhale images after deformable image registration. All VIs were deformably registered to their corresponding planning CT and normalized to account for differences in breathing effort, thus facilitating image comparison in terms of (i) voxelwise Spearman correlations, (ii) mean image differences, and (iii) gamma pass rates for all interfraction and intrafraction VI pairs. For the side of the lung ipsilateral to the tumor, we applied two-sided t-tests to determine whether interfraction VI pairs were more different than intrafraction VI pairs. RESULTS: The (mean ± standard deviation) Spearman correlation for interfraction VI pairs was r̄(Inter)=0.52±0.25, which was significantly lower than for intrafraction pairs (r̄(Intra)=0.67±0.20, p = 0.0002). Conversely, mean absolute ventilation differences were larger for interfraction pairs than for intrafraction pairs, with |ΔV̄(Inter)|=0.42±0.65 and |ΔV̄(Intra)|=0.32±0.53, respectively (p < 10(-15)). Applying a gamma analysis with ventilation/distance tolerance of 25%/10 mm, we observed mean pass rate of (69% ± 20%) for interfraction VIs, which was significantly lower compared to intrafraction pairs (80% ± 15%, with p ∼ 0.0003). Compared to the first day scans, all patients experienced at least one subsequent change in median ipsilateral ventilation ≥10%. Patients experienced both positive and negative ventilation changes throughout treatment, with the maximum change occurring at different weeks for different patients. CONCLUSIONS: The authors' data support the hypothesis that interfraction ventilation changes are larger than intrafraction ventilation changes for LA-NSCLC patients over a course of conventional lung cancer radiation therapy. Longitudinal ventilation changes are observed to be highly patient-dependent, supporting a possible role for adaptive ventilation guidance based on repeat 4D-CBCT VIs. We anticipate that future improvement of 4D-CBCT image reconstruction algorithms will improve the capability of 4D-CBCT VI to resolve interfraction ventilation changes.

Improving thoracic four-dimensional cone-beam CT reconstruction with anatomical-adaptive image regularization (AAIR).

Total-variation (TV) minimization reconstructions can significantly reduce noise and streaks in thoracic four-dimensional cone-beam computed tomography (4D CBCT) images compared to the Feldkamp-Davis-Kress (FDK) algorithm currently used in practice. TV minimization reconstructions are, however, prone to over-smoothing anatomical details and are also computationally inefficient. The aim of this study is to demonstrate a proof of concept that these disadvantages can be overcome by incorporating the general knowledge of the thoracic anatomy via anatomy segmentation into the reconstruction. The proposed method, referred as the anatomical-adaptive image regularization (AAIR) method, utilizes the adaptive-steepest-descent projection-onto-convex-sets (ASD-POCS) framework, but introduces an additional anatomy segmentation step in every iteration. The anatomy segmentation information is implemented in the reconstruction using a heuristic approach to adaptively suppress over-smoothing at anatomical structures of interest. The performance of AAIR depends on parameters describing the weighting of the anatomy segmentation prior and segmentation threshold values. A sensitivity study revealed that the reconstruction outcome is not sensitive to these parameters as long as they are chosen within a suitable range. AAIR was validated using a digital phantom and a patient scan and was compared to FDK, ASD-POCS and the prior image constrained compressed sensing (PICCS) method. For the phantom case, AAIR reconstruction was quantitatively shown to be the most accurate as indicated by the mean absolute difference and the structural similarity index. For the patient case, AAIR resulted in the highest signal-to-noise ratio (i.e. the lowest level of noise and streaking) and the highest contrast-to-noise ratios for the tumor and the bony anatomy (i.e. the best visibility of anatomical details). Overall, AAIR was much less prone to over-smoothing anatomical details compared to ASD-POCS and did not suffer from residual noise/streaking and motion blur migrated from the prior image as in PICCS. AAIR was also found to be more computationally efficient than both ASD-POCS and PICCS, with a reduction in computation time of over 50% compared to ASD-POCS. The use of anatomy segmentation was, for the first time, demonstrated to significantly improve image quality and computational efficiency for thoracic 4D CBCT reconstruction. Further developments are required to facilitate AAIR for practical use.

Optimizing 4DCBCT projection allocation to respiratory bins.

4D cone beam computed tomography (4DCBCT) is an emerging image guidance strategy used in radiotherapy where projections acquired during a scan are sorted into respiratory bins based on the respiratory phase or displacement. 4DCBCT reduces the motion blur caused by respiratory motion but increases streaking artefacts due to projection under-sampling as a result of the irregular nature of patient breathing and the binning algorithms used. For displacement binning the streak artefacts are so severe that displacement binning is rarely used clinically. The purpose of this study is to investigate if sharing projections between respiratory bins and adjusting the location of respiratory bins in an optimal manner can reduce or eliminate streak artefacts in 4DCBCT images. We introduce a mathematical optimization framework and a heuristic solution method, which we will call the optimized projection allocation algorithm, to determine where to position the respiratory bins and which projections to source from neighbouring respiratory bins. Five 4DCBCT datasets from three patients were used to reconstruct 4DCBCT images. Projections were sorted into respiratory bins using equispaced, equal density and optimized projection allocation. The standard deviation of the angular separation between projections was used to assess streaking and the consistency of the segmented volume of a fiducial gold marker was used to assess motion blur. The standard deviation of the angular separation between projections using displacement binning and optimized projection allocation was 30%-50% smaller than conventional phase based binning and 59%-76% smaller than conventional displacement binning indicating more uniformly spaced projections and fewer streaking artefacts. The standard deviation in the marker volume was 20%-90% smaller when using optimized projection allocation than using conventional phase based binning suggesting more uniform marker segmentation and less motion blur. Images reconstructed using displacement binning and the optimized projection allocation algorithm were clearer, contained visibly fewer streak artefacts and produced more consistent marker segmentation than those reconstructed with either equispaced or equal-density binning. The optimized projection allocation algorithm significantly improves image quality in 4DCBCT images and provides, for the first time, a method to consistently generate high quality displacement binned 4DCBCT images in clinical applications.