Geometric evaluation and quantifying dosimetric impact of diverse deformable image registration algorithms on abdomen images with biomechanically modeled deformations.

PURPOSE: Deformable image registration (DIR) has been increasingly used in radiation therapy (RT). The accuracy of DIR algorithms and how it impacts on the RT plan dosimetrically were examined in our study for abdominal sites using biomechanically modeled deformations. METHODS: Five pancreatic cancer patients were enrolled in this study. Following the guidelines of AAPM TG-132, a patient-specific quality assurance (QA) workflow was developed to evaluate DIR for the abdomen using the TG-132 recommended virtual simulation software ImSimQA (Shrewsbury, UK). First, the planning CT was deformed to simulate respiratory motion using the embedded biomechanical model in ImSimQA. Additionally, 5 mm translational motion was added to the stomach, duodenum, and small bowel. The original planning CT and the deformed CT were then imported into Eclipse and MIM to perform DIR. The output displacement vector fields (DVFs) were compared with the ground truth from ImSimQA. Furthermore, the original treatment plan was recalculated on the ground-truth deformed CT and the deformed CT (with Eclipse and MIM DVF). The dose errors were calculated on a voxel-to-voxel basis. RESULTS: Data analysis comparing DVF from Eclipse versus MIM show the average mean DVF magnitude errors of 2.8 ± 1.0  versus 1.1 ± 0.7 mm for stomach and duodenum, 5.2 ± 4.0  versus 2.5 ± 1.0 mm for small bowel, and 4.8 ± 4.1  versus 2.7 ± 1.1 mm for the gross tumor volume (GTV), respectively, across all patients. The mean dose error on stomach+duodenum and small bowel were 2.3 ± 0.6% for Eclipse, and 1.0 ± 0.3% for MIM. As the DIR magnitude error increases, the dose error range increase, for both Eclipse and MIM. CONCLUSION: In our study, an initial assessment was conducted to evaluate the accuracy of DIR and its dosimetric impact on radiotherapy. A patient-specific DIR QA workflow was developed for pancreatic cancer patients. This workflow exhibits promising potential for future implementation as a clinical workflow.

Comparative study of SRS end-to-end QA processes of a diode array device and an anthropomorphic phantom loaded with GafChromic XD film.

PURPOSE: End-to-end testing (E2E) is a necessary process for assessing the readiness of the stereotactic radiosurgery (SRS) program and annual QA of an SRS system according to the AAPM MPPG 9a. This study investigates the differences between using a new SRS MapCHECK (SRSMC) system and an anthropomorphic phantom film-based system in a large network with different SRS delivery techniques. METHODS AND MATERIALS: Three SRS capable Linacs (Varian Medical Systems, Palo Alto, CA) at three different regional sites were chosen to represent a hospital network, a Trilogy with an M120 multi-leaf collimator (MLC), a TrueBeam with an M120 MLC, and a TrueBeam Stx with an HD120 MLC. An anthropomorphic STEEV phantom (CIRS, Norfolk, VA) and a phantom/diode array: StereoPHAN/SRSMC (Sun Nuclear, Melbourne, FL) were CT scanned at each site. The new STV-PHANTOM EBT-XD films (Ashland, Bridgewater, NJ) were used. Six plans with various complexities were measured with both films and SRSMC in the StereoPHAN to establish their dosimetric correlations. Three SRS cranial plans with a total of sixteen fields using dynamic conformal arc and volumetric-modulated arc therapy, with 1-4 targets, were planned with Eclipse v15.5 treatment planning system (TPS) using a custom SRS beam model for each machine. The dosimetric and localization accuracy were compared. The time of analysis for the two systems by three teams of physicists was also compared to assess the throughput efficiency. RESULTS: The correlations between films and SRSMC were found to be 0.84 (p = 0.03) and 0.16 (p = 0.76) for γ (3%, 1 mm) and γ (3%, 2 mm), respectively. With film, the local dose differences (ΔD) relative to the average dose within the 50% isodose line from the three sites were found to be -3.2%-3.7%. The maximum localization errors (Elocal ) were found to be within 0.5 ± 0.2 mm. With SRSMC, the ΔD was found to be within 5% of the TPS calculation. Elocal were found to be within 0.7 to 1.1 ± 0.4 mm for TrueBeam and Trilogy, respectively. Comparing with film, an additional uncertainty of 0.7 mm was found with SRSMC. The delivery and analysis times were found to be 6 and 2 h for film and SRSMC, respectively. CONCLUSIONS: The SRS MapCHECK agrees dosimetrically with the films within measurement uncertainties. However, film dosimetry shows superior sub-millimeter localization resolving power for the MPPG 9a implementation.

Implementation and validation of an in-house geometry optimization software for SRS VMAT planning of multiple cranial metastases.

PURPOSE: The implementation and evaluation of an in-house developed geometry optimization (GO) software are described. The GO script provides optimal lesion clustering, isocenter placement, and collimator angle of each arc for cranial multi-lesion stereotactic radiosurgery (SRS) volumetric modulated arc therapy (VMAT) planning. MATERIALS AND METHODS: An Eclipse-plugin program was developed to facilitate automatic plan geometry generation for multiple metastases SRS VMAT plans. A mixed, semi-supervised exhaustive and k-means clustering method is used to group lesions and place isocenters. The sum of squared euclidean distance (SSED) and the boundaries of lesions' projection from beams' eye view are used as supervised parameters to determine the optimal isocenter numbers. The collimator angle is optimized by minimizing the sum of the MLC opening area from all gantry angles for each arc. In all, 10 clinical cases treated during 2016-2017 were compared to plan quality of GO script generated plans. Paddick gradient index (GI), conformity index (CI), and local brain volume receiving 12 Gy (local V12 Gy) around each lesion were compared. RESULT: For four cases, the number of isocenters was reduced in the GO plans. For four other cases, the GO plans had the same number of isocenters as their corresponding clinical plans but with different lesion grouping. The GO plans had significantly lower GI (4.1 ± 1.0 vs 4.4 ± 0.9, P < 0.0001) and local V12 Gy (5.1 ± 4.2 vs 5.5 ± 4.3 in cm3 , P < 0.0001), but not significantly different mean normal brain dose or CI. The volume of normal brain receiving ≥6 Gy was significantly lower in the GO plans. The total time to run the GO script for each case was <2 min. CONCLUSION: The GO software automates lesion grouping, isocenter placement, and the collimator angles for SRS VMAT planning. When tested on 10 cases, the GO script resulted in improved plan quality and shorter planning time when compared to the clinical SRS VMAT plans.

Geometric dose prediction model for hemithoracic intensity-modulated radiation therapy in mesothelioma patients with two intact lungs.

The presence of two intact lungs makes it challenging to reach a tumoricidal dose with hemithoracic pleural intensity-modulated radiation therapy (IMRT) in patients with malignant pleural mesothelioma (MPM) who underwent pleurectomy/decortications or have unresectable disease. We developed an anatomy-based model to predict attainable prescription dose before starting optimization. Fifty-six clinically delivered IMRT plans were analyzed regarding correlation of prescription dose and individual and total lung volumes, planning target volume (PTV), ipsilateral normal lung volume and ratios: contralateral/ipsilateral lung (CIVR); contralateral lung/PTV (CPVR); ipsilateral lung /PTV (IPVR); ipsilateral normal lung /total lung (INTLVR); ipsilateral normal lung/PTV (INLPVR). Spearman's rank correlation and Fisher's exact test were used. Correlation between mean ipsilateral lung dose (MILD) and these volume ratios and between prescription dose and single lung mean doses were studied. The prediction models were validated in 23 subsequent MPM patients. CIVR showed the strongest correlation with dose (R=0.603,p<0.001) and accurately predicted prescription dose in the validation cases. INLPVR and MILD as well as MILD and prescription dose were significantly correlated (R=-0.784,p<0.001 and R=0.554,p<0.001, respectively) in the training and validation cases. Parameters obtainable directly from planning scan anatomy predict achievable prescription doses for hemithoracic IMRT treatment of MPM patients with two intact lungs. PACS number(s): 87.55.de, 87.55.dk.

Examining feature extraction and classification modules in machine learning for diagnosis of low-dose computed tomographic screening-detected in vivo lesions.

Purpose: Medical imaging-based machine learning (ML) for computer-aided diagnosis of in vivo lesions consists of two basic components or modules of (i) feature extraction from non-invasively acquired medical images and (ii) feature classification for prediction of malignancy of lesions detected or localized in the medical images. This study investigates their individual performances for diagnosis of low-dose computed tomography (CT) screening-detected lesions of pulmonary nodules and colorectal polyps. Approach: Three feature extraction methods were investigated. One uses the mathematical descriptor of gray-level co-occurrence image texture measure to extract the Haralick image texture features (HFs). One uses the convolutional neural network (CNN) architecture to extract deep learning (DL) image abstractive features (DFs). The third one uses the interactions between lesion tissues and X-ray energy of CT to extract tissue-energy specific characteristic features (TFs). All the above three categories of extracted features were classified by the random forest (RF) classifier with comparison to the DL-CNN method, which reads the images, extracts the DFs, and classifies the DFs in an end-to-end manner. The ML diagnosis of lesions or prediction of lesion malignancy was measured by the area under the receiver operating characteristic curve (AUC). Three lesion image datasets were used. The lesions' tissue pathological reports were used as the learning labels. Results: Experiments on the three datasets produced AUC values of 0.724 to 0.878 for the HFs, 0.652 to 0.965 for the DFs, and 0.985 to 0.996 for the TFs, compared to the DL-CNN of 0.694 to 0.964. These experimental outcomes indicate that the RF classifier performed comparably to the DL-CNN classification module and the extraction of tissue-energy specific characteristic features dramatically improved AUC value. Conclusions: The feature extraction module is more important than the feature classification module. Extraction of tissue-energy specific characteristic features is more important than extraction of image abstractive and characteristic features.

Radiotherapy Dose Accumulation Routine (RADAR)- A novel dose accumulation script with built-in uncertainty.

PURPOSE: To incorporate uncertainty into dose accumulation for reirradiation. METHODS: The RADAR script for the Eclipse treatment planning system (Varian Medical Systems, Palo Alto, CA) is described and the voxel-wise ellipsoid search algorithm is introduced as a means of incorporating uncertainty. RADAR is first demonstrated on a test patient reirradiated to the spine illustrating the effect of the uncertainty algorithm. A summary of initial evaluation testing by 11 users, each of whom ran a separate spine reirradiation case, follows. Finally, RADAR run times are reported for different conditions. RESULTS: In the demonstration case in which a 3mm ellipsoid search was used, maximum RADAR 2 Gy equivalent (EQD2) accumulated spinal cord dose increased from 7244 cGy to 12689 cGy because the ellipsoid search pulled dose from closer to the adjacent target structure. When the ellipsoid search was restricted to voxels within the spinal cord, the maximum accumulated cord dose was reduced to 6523 cGy and did not exceed the sum of the maximum EQD2 spinal cord doses of the individual plans (6730 cGy). In the evaluation cases, the RADAR EQD2 maximum dose for the spinal cord increased an average of 31.6% with uncertainty applied compared to a conventional dose accumulation and decreased an average of 16.7% compared to a conventional dose accumulation when the uncertainty calculation was restricted to voxels within the spinal cord. RADAR run times depend on the number of plans being added and the type of uncertainty being used. CONCLUSION: RADAR offers a novel way to directly account for uncertainty in dose accumulation by means of a voxel-wise ellipsoid search algorithm. EQD2 dose accumulation with and without dose discounts is also available.

A Potential Pitfall and Clinical Solutions in Surface-Guided Deep Inspiration Breath Hold Radiation Therapy for Left-Sided Breast Cancer.

Purpose: Deep inspiration breath hold (DIBH) is an effective technique to spare the heart in treating left-sided breast cancer. Surface-guided radiation therapy (SGRT) is increasingly applied in DIBH setup and motion monitoring. Patient-specific breathing behavior, either thoracically driven or abdominally driven (A-DIBH), should be unaltered, online identified, and monitored accordingly to ensure reproducible heart-sparing treatment. Methods and Materials: Sixty patients with left-sided breast cancer treated with SGRT were analyzed: 20 A-DIBH patients with vertical chest elevation (VCE ≤ 5 mm) were prospectively identified, and 40 control patients were retrospectively and randomly selected for comparison. At simulation, both free-breathing (FB) and DIBH computed tomography (CT) were acquired, guided by a motion surrogate placed around the xiphoid process. For SGRT treatment setups, the region of interest (ROI) was defined on the CT chest surface, and the surrogate-based setup was a backup. For all 60 patients, the VCE was measured as the average of the FB-to-DIBH elevations at the breast and xiphoid process, together with abdominal elevation. In the 40-patient control group, A-DIBH patients (VCE ≤ 5 mm) were identified. Of the 20 A-DIBH patients, 10 were treated with volumetric modulated arc therapy plans, and 10 patients were treated with tangent plans. Clinical DIBH plans were recalculated on FB CT to compare maximum dose (DMax), 5% of the maximum dose (D5%), mean dose (DMean), and V30Gy, V20Gy, and V5Gy of the heart and lungs and their significance. Results: In the 20 A-DIBH patients, VCE = 3 ± 2 mm, surrogate motion (9 ± 6 mm), and abdomen motion of 14 ± 5 mm are found. Heart dose reduction from FB to DIBH is significant (P < .01): ∆DMax = -8.4 ± 9.8 Gy, ∆D5% = -2.4 ± 4.4 Gy, and ∆DMean = -0.6 ± 0.9 Gy. Six out of 40 control patients (15%) are found to have VCE ≤ 5 mm. Conclusions: A-DIBH (VCE ≤ 5 mm) patient population is significant (15%), and they should be identified in the SGRT workflow and monitored accordingly. A new abdominal ROI or an abdominal surrogate should be used instead of the conventional chest-only ROI. Patient-specific DIBH should be preserved for higher reproducibility to ensure heart sparing.

MM-GLCM-CNN: A multi-scale and multi-level based GLCM-CNN for polyp classification.

Distinguishing malignant from benign lesions has significant clinical impacts on both early detection and optimal management of those early detections. Convolutional neural network (CNN) has shown great potential in medical imaging applications due to its powerful feature learning capability. However, it is very challenging to obtain pathological ground truth, addition to collected in vivo medical images, to construct objective training labels for feature learning, leading to the difficulty of performing lesion diagnosis. This is contrary to the requirement that CNN algorithms need a large number of datasets for the training. To explore the ability to learn features from small pathologically-proven datasets for differentiation of malignant from benign polyps, we propose a Multi-scale and Multi-level based Gray-level Co-occurrence Matrix CNN (MM-GLCM-CNN). Specifically, instead of inputting the lesions' medical images, the GLCM, which characterizes the lesion heterogeneity in terms of image texture characteristics, is fed into the MM-GLCN-CNN model for the training. This aims to improve feature extraction by introducing multi-scale and multi-level analysis into the construction of lesion texture characteristic descriptors (LTCDs). To learn and fuse multiple sets of LTCDs from small datasets for lesion diagnosis, we further propose an adaptive multi-input CNN learning framework. Furthermore, an Adaptive Weight Network is used to highlight important information and suppress redundant information after the fusion of the LTCDs. We evaluated the performance of MM-GLCM-CNN by the area under the receiver operating characteristic curve (AUC) merit on small private lesion datasets of colon polyps. The AUC score reaches 93.99% with a gain of 1.49% over current state-of-the-art lesion classification methods on the same dataset. This gain indicates the importance of incorporating lesion characteristic heterogeneity for the prediction of lesion malignancy using small pathologically-proven datasets.

Deep learning driven predictive treatment planning for adaptive radiotherapy of lung cancer.

BACKGROUND AND PURPOSE: To develop a novel deep learning algorithm of sequential analysis, Seq2Seq, for predicting weekly anatomical changes of lung tumor and esophagus during definitive radiotherapy, incorporate the potential tumor shrinkage into a predictive treatment planning paradigm, and improve the therapeutic ratio. METHODS AND MATERIALS: Seq2Seq starts with the primary tumor and esophagus observed on the planning CT to predict their geometric evolution during radiotherapy on a weekly basis, and subsequently updates the predictions with new snapshots acquired via weekly CBCTs. Seq2Seq is equipped with convolutional long short term memory to analyze the spatial-temporal changes of longitudinal images, trained and validated using a dataset including sixty patients. Predictive plans were optimized according to each weekly prediction and made ready for weekly deployment to mitigate the clinical burden of online weekly replanning. RESULTS: Seq2Seq tracks structural changes well: DICE between predicted and actual weekly tumor and esophagus were (0.83 ± 0.10, 0.79 ± 0.14, 0.78 ± 0.12, 0.77 ± 0.12, 0.75 ± 0.12, 0.71 ± 0.17), and (0.72 ± 0.16, 0.73 ± 0.11, 0.75 ± 0.08, 0.74 ± 0.09, 0.72 ± 0.14, 0.71 ± 0.14), respectively, while the average Hausdorff distances were within 2 mm. Evaluating dose to the actual weekly tumor and esophagus, a 4.2 Gy reduction in esophagus mean dose while maintaining 60 Gy tumor coverage was achieved with the predictive weekly plans, compared to the plan optimized using the initial tumor and esophagus alone, primarily due to noticeable tumor shrinkage during radiotherapy. CONCLUSION: It is feasible to predict the longitudinal changes of tumor and esophagus with the Seq2Seq, which could lead to improving the efficiency and effectiveness of lung adaptive radiotherapy.

A VMAT planning technique for locally advanced breast cancer patients with expander or implant reconstructions requiring comprehensive postmastectomy radiation therapy.

Locally advanced breast cancer patients with expander or implant reconstructions who require comprehensive postmastectomy radiotherapy (PMRT) can pose unique treatment planning challenges. Traditional 3D conformal radiation techniques often result in large dose inhomogeneity throughout the treatment volumes, inadequate target coverage, or excessive normal tissue doses. We have developed a volumetric modulated arc therapy (VMAT) planning technique without entering through the ipsilateral arm that produced adequate target volume coverage, excellent homogeneity throughout the target volume, and acceptable doses to the normal structures. Twenty left-sided and 10 right-sided patients with either ipsilateral or bilateral permanent implants or tissue expanders who received comprehensive PMRT between October 2014 and February 2016 were included in this study. Ten left-sided cases used deep inspiration breath hold (DIBH) technique, and others used free breathing (FB). Planning target volume (PTV) included chestwall, internal mammary nodes (IMNs), supraclavicular, and axillary lymph nodes. A VMAT plan using 4 or 5 partial arcs with 6 MV photon beam avoiding entering through the ipsilateral arm was generated for each patient. Prescription dose was 50 Gy in 25 fractions. PTV coverage, maximum depth of IMNs, dose homogeneity and dose to the heart, lungs, thyroid, contralateral intact breast or implant, liver, stomach, left anterior descending artery, ipsilateral brachial plexus, esophagus, spinal cord, and total MU were evaluated. PTV D95% (Gy) was 49.6 ± 0.9, 48.7 ± 0.9, and 49.5 ± 1.1; PTV D05% (Gy) was 55.7 ± 0.6, 55.1 ± 1.4, and 55.0 ± 0.7; maximum depth of IMNs (cm) was 4.3 ± 0.9, 4.6 ± 1.1, and 4.9 ± 2.3; ipsilateral lung, V20Gy (%) was 29.0 ± 2.1, 28.8 ± 2.5, and 27.5 ± 3.4; heart mean dose (Gy) was 4.2 ± 0.4, 7.5 ± 1.1, and 6.6 ± 0.8 for right-sided FB, left-sided FB, and left-sided DIBH cases, respectively. D95% of IMNs all received 100% prescription dose. The maximum dose (Gy) to the left anterior descending artery was 33.8 ± 11.7 for left-sided FB and 31.4 ± 7.3 for left-sided DIBH. VMAT technique avoiding ipsilateral arm can produce acceptable clinical plans for locally advanced breast cancer patients with expander or implant reconstructions receiving comprehensive PMRT.

A geometric atlas to predict lung tumor shrinkage for radiotherapy treatment planning.

To develop a geometric atlas that can predict tumor shrinkage and guide treatment planning for non-small-cell lung cancer. To evaluate the impact of the shrinkage atlas on the ability of tumor dose escalation. The creation of a geometric atlas included twelve patients with lung cancer who underwent both planning CT and weekly CBCT for radiotherapy planning and delivery. The shrinkage pattern from the original pretreatment to the residual posttreatment tumor was modeled using a principal component analysis, and used for predicting the spatial distribution of the residual tumor. A predictive map was generated by unifying predictions from each individual patient in the atlas, followed by correction for the tumor's surrounding tissue distribution. Sensitivity, specificity, and accuracy of the predictive model for classifying voxels inside the original gross tumor volume were evaluated. In addition, a retrospective study of predictive treatment planning (PTP) escalated dose to the predicted residual tumor while maintaining the same level of predicted complication rates for a clinical plan delivering uniform dose to the entire tumor. The effect of uncertainty on the predictive model's ability to escalate dose was also evaluated. The sensitivity, specificity and accuracy of the predictive model were 0.73, 0.76, and 0.74, respectively. The area under the receiver operating characteristic curve for voxel classification was 0.87. The Dice coefficient and mean surface distance between the predicted and actual residual tumor averaged 0.75, and 1.6 mm, respectively. The PTP approach allowed elevation of PTV D95 and mean dose to the actual residual tumor by 6.5 Gy and 10.4 Gy, respectively, relative to the clinical uniform dose approach. A geometric atlas can provide useful information on the distribution of resistant tumors and effectively guide dose escalation to the tumor without compromising the organs at risk complications. The atlas can be further refined by using more patient data sets.

A study on the dose distributions in various materials from an Ir-192 HDR brachytherapy source.

Dose distributions of (192)Ir HDR brachytherapy in phantoms simulating water, bone, lung tissue, water-lung and bone-lung interfaces using the Monte Carlo codes EGS4, FLUKA and MCNP4C are reported. Experiments were designed to gather point dose measurements to verify the Monte Carlo results using Gafchromic film, radiophotoluminescent glass dosimeter, solid water, bone, and lung phantom. The results for radial dose functions and anisotropy functions in solid water phantom were consistent with previously reported data (Williamson and Li). The radial dose functions in bone were affected more by depth than those in water. Dose differences between homogeneous solid water phantoms and solid water-lung interfaces ranged from 0.6% to 14.4%. The range between homogeneous bone phantoms and bone-lung interfaces was 4.1% to 15.7%. These results support the understanding in dose distribution differences in water, bone, lung, and their interfaces. Our conclusion is that clinical parameters did not provide dose calculation accuracy for different materials, thus suggesting that dose calculation of HDR treatment planning systems should take into account material density to improve overall treatment quality.