Evaluation of measures for assessing time-saving of automatic organ-at-risk segmentation in radiotherapy

•

Automatic delineation software shows promising results in terms of time-saving.

•

Standard geometry measures do not have a high correlation with delineation time.

•

New evaluation measures were introduced: added path length (APL) and surface DSC.

•

(Added) path length showed the highest correlation with time-recordings.

•

This makes APL the most representative measure for clinical usefulness.

Background and purpose

In radiotherapy, automatic organ-at-risk segmentation algorithms allow faster delineation times, but clinically relevant contour evaluation remains challenging. Commonly used measures to assess automatic contours, such as volumetric Dice Similarity Coefficient (DSC) or Hausdorff distance, have shown to be good measures for geometric similarity, but do not always correlate with clinical applicability of the contours, or time needed to adjust them. This study aimed to evaluate the correlation of new and commonly used evaluation measures with time-saving during contouring.

Materials and methods

Twenty lung cancer patients were used to compare user-adjustments after atlas-based and deep-learning contouring with manual contouring. The absolute time needed (s) of adjusting the auto-contour compared to manual contouring was recorded, from this relative time-saving (%) was calculated. New evaluation measures (surface DSC and added path length, APL) and conventional evaluation measures (volumetric DSC and Hausdorff distance) were correlated with time-recordings and time-savings, quantified with the Pearson correlation coefficient, R.

Results

The highest correlation (R = 0.87) was found between APL and absolute adaption time. Lower correlations were found for APL with relative time-saving (R = −0.38), for surface DSC with absolute adaption time (R = −0.69) and relative time-saving (R = 0.57). Volumetric DSC and Hausdorff distance also showed lower correlation coefficients for absolute adaptation time (R = −0.32 and 0.64, respectively) and relative time-saving (R = 0.44 and −0.64, respectively).

Conclusion

Surface DSC and APL are better indicators for contour adaptation time and time-saving when using auto-segmentation and provide more clinically relevant and better quantitative measures for automatically-generated contour quality, compared to commonly-used geometry-based measures.

Hybrid deep learning segmentation models for atherosclerotic plaque in internal carotid artery B-mode ultrasound

The automated and accurate carotid plaque segmentation in B-mode ultrasound (US) is an essential part of stroke risk stratification. Previous segmented methods used AtheroEdge™ 2.0 (AtheroPoint™, Roseville, CA) for the common carotid artery (CCA). This study focuses on automated plaque segmentation in the internal carotid artery (ICA) using solo deep learning (SDL) and hybrid deep learning (HDL) models. The methodology consists of a novel design of 10 types of SDL/HDL models (AtheroEdge™ 3.0 systems (AtheroPoint™, Roseville, CA) with a depth of four layers each. Five of the models use cross-entropy (CE)-loss, and the other five models use Dice similarity coefficient (DSC)-loss functions derived from UNet, UNet+, SegNet, SegNet-UNet, and SegNet-UNet+. The K10 protocol (Train:Test:90%:10%) was applied for all 10 models for training and predicting (segmenting) the plaque region, which was then quantified to compute the plaque area in mm². Further, the data augmentation effect was analyzed. The database consisted of 970 ICA B-mode US scans taken from 99 moderate to high-risk patients. Using the difference area threshold of 10 mm² between ground truth (GT) and artificial intelligence (AI), the area under the curve (AUC) values were 0.91, 0.911, 0.908, 0.905, and 0.898, all with a p-value of <0.001 (for CE-loss models) and 0.883, 0.889, 0.905, 0.889, and 0.907, all with a p-value of <0.001 (for DSC-loss models). The correlations between the AI-based plaque area and GT plaque area were 0.98, 0.96, 0.97, 0.98, and 0.97, all with a p-value of <0.001 (for CE-loss models) and 0.98, 0.98, 0.97, 0.98, and 0.98 (for DSC-loss models). Overall, the online system performs plaque segmentation in less than 1 s. We validate our hypothesis that HDL and SDL models demonstrate comparable performance. SegNet-UNet was the best-performing hybrid architecture.

Registration of in vivo prostate MRI and pseudo-whole mount histology using Local Affine Transformations guided by Internal Structures (LATIS)

PURPOSE

To present a novel registration approach called LATIS (Local Affine Transformation guided by Internal Structures) for coregistering post prostatectomy pseudo-whole mount (PWM) pathological sections with in vivo MRI (magnetic resonance imaging) images.

MATERIALS AND METHODS

Thirty-five patients with biopsy-proven prostate cancer were imaged at 3T with an endorectal coil. Excised prostate specimens underwent quarter mount step-section pathologic processing, digitization, annotation, and assembly into a PWM. Manually annotated macro-structures on both pathology and MRI were used to assist registration using a relaxed local affine transformation approximation. Registration accuracy was assessed by calculation of the Dice similarity coefficient (DSC) between transformed and target capsule masks and least-square distance between transformed and target landmark positions.

RESULTS

LATIS registration resulted in a DSC value of 0.991 ± 0.004 and registration accuracy of 1.54 ± 0.64 mm based on identified landmarks common to both datasets. Image registration performed without the use of internal structures led to an 87% increase in landmark-based registration error. Derived transformation matrices were used to map regions of pathologically defined disease to MRI.

CONCLUSION

LATIS was used to successfully coregister digital pathology with in vivo MRI to facilitate improved correlative studies between pathologically identified features of prostate cancer and multiparametric MRI.

Evaluation of the correlation between prostatic displacement and rectal deformation using the Dice similarity coefficient of the rectum

To estimate the relationship between the three-dimensional (3D) displacement error of the prostate and rectal deformation for reduction of deviation between the planned and treatment dose, using multiple acquisition planning CT (MPCT) and the Dice similarity coefficient (DSC) for rectal deformation for treatment of patients with prostate cancer. The 3D displacement error between the pelvic bone and a matching fiducial marker was calculated using MPCT in 24 patients who underwent prostate volumetric-modulated arc therapy for prostate cancer. We calculated the 3D displacement error between the pelvic bone and a matching fiducial marker on MPCT. The correlation of the 3D displacement error with the DSC of the rectum, calculated from MPCT images, was evaluated based on deformable image registration. The 3D displacement error of the prostate showed a slight correlation between MPCT and cone-beam computed tomography (adjusted r² = 0.241). The 3D displacement error, based on the pelvic bone and a fiducial marker on MPCT images, showed a moderate correlation with the DSC of the rectum (adjusted r² = 0.645) and was improved by a mean of 3.94 mm, based on MPCT, during the treatment period. The 3D displacement error on MPCT correlates with the 3D displacement error of daily cone-beam computed tomography; optimal selection of MPCT can potentially facilitate on-board setup of prostate patients to enable more accurate radiotherapy. The advance information of the 3D displacement error and rectal deformation is useful for optimal planning CT that can minimize the deviation between the planned dose and the treatment dose in patients receiving treatment for prostate cancer.

Are multi-contrast magnetic resonance images necessary for segmenting multiple sclerosis brains? A large cohort study based on deep learning

BACKGROUND

Magnetic resonance images with multiple contrasts or sequences are commonly used for segmenting brain tissues, including lesions, in multiple sclerosis (MS). However, acquisition of images with multiple contrasts increases the scan time and complexity of the analysis, possibly introducing factors that could compromise segmentation quality.

OBJECTIVE

To investigate the effect of various combinations of multi-contrast images as input on the segmented volumes of gray (GM) and white matter (WM), cerebrospinal fluid (CSF), and lesions using a deep neural network.

METHODS

U-net, a fully convolutional neural network was used to automatically segment GM, WM, CSF, and lesions in 1000 MS patients. The input to the network consisted of 15 combinations of FLAIR, T1-, T2-, and proton density-weighted images. The Dice similarity coefficient (DSC) was evaluated to assess the segmentation performance. For lesions, true positive rate (TPR) and false positive rate (FPR) were also evaluated. In addition, the effect of lesion size on lesion segmentation was investigated.

RESULTS

Highest DSC was observed for all the tissue volumes, including lesions, when the input was combination of all four image contrasts. All other input combinations that included FLAIR also provided high DSC for all tissue classes. However, the quality of lesion segmentation showed strong dependence on the input images. The DSC and TPR values for inputs with the four contrast combination and FLAIR alone were very similar, but FLAIR showed a moderately higher FPR for lesion size <100 μl. For lesions smaller than 20 μl all image combinations resulted in poor performance. The segmentation quality improved with lesion size.

CONCLUSIONS

Best performance for segmented tissue volumes was obtained with all four image contrasts as the input, and comparable performance was attainable with FLAIR only as the input, albeit with a moderate increase in FPR for small lesions. This implies that acquisition of only FLAIR images provides satisfactory tissue segmentation. Lesion segmentation was poor for very small lesions and improved rapidly with lesion size.

Changes in Tumor Volumes and Spatial Locations Relative to Normal Tissues During Cervical Cancer Radiotherapy Assessed by Cone Beam Computed Tomography

Purpose:

To assess changes in the volumes and spatial locations of tumors and surrounding organs by cone beam computed tomography during treatment for cervical cancer.

Materials and Methods:

Sixteen patients with cervical cancer had intensity-modulated radiotherapy and off-line cone beam computed tomography during chemotherapy and/or radiation therapy. The gross tumor volume (GTV-T) and clinical target volumes (CTVs) were contoured on the planning computed tomography and weekly cone beam computed tomography image, and changes in volumes and spatial locations were evaluated using the volume difference method and Dice similarity coefficients.

Results:

The GTV-T was 79.62 cm³ at prior treatment (0f) and then 20.86 cm³ at the end of external-beam chemoradiation. The clinical target volume changed slightly from 672.59 cm³ to 608.26 cm³, and the uterine volume (CTV-T) changed slightly from 83.72 cm³ to 80.23 cm³. There were significant differences in GTV-T and CTV-T among the different groups (P < .001), but the clinical target volume was not significantly different in volume (P > .05). The mean percent volume changes ranged from 23.05% to 70.85% for GTV-T, 4.71% to 6.78% for CTV-T, and 5.84% to 9.59% for clinical target volume, and the groups were significantly different (P < .05). The Dice similarity coefficient of GTV-T decreased during the course of radiation therapy (P < .001). In addition, there were significant differences in GTV-T among different groups (P < .001), and changes in GTV-T correlated with the radiotherapy (P < .001). There was a negative correlation between volume change rate (DV) and Dice similarity coefficient in the GTV-T and organs at risk (r < 0; P < .05).

Conclusion:

The volume, volume change rate, and Dice similarity coefficient of GTV-T were all correlated with increase in radiation treatment. Significant variations in tumor regression and spatial location occurred during radiotherapy for cervical cancer. Adaptive radiotherapy approaches are needed to improve the treatment accuracy for cervical cancer.

Clinical relevance of metal artefact reduction in computed tomography (iMAR) in the pelvic and head and neck region: Multi-institutional contouring study of gross tumour volumes and organs at risk on clinical cases

INTRODUCTION

Artefacts caused by dental implants and hip replacements may impede target volume definition and dose calculation accuracy. The iterative metal artefact reduction (iMAR) algorithm can provide a solution for this problem. The present study compares delineation of gross tumour volumes (GTVs) and organs at risk (OARs) in the pelvic and the head and neck (H & N) regions using computed tomography (CT) with and without iMAR, and thus the practical applicability of iMAR for routine clinical use.

METHODS

The native planning CT and CT-iMAR data of two typical clinical cases with image-distorting artefacts were used for multi-institutional contouring and analysis using the Dice similarity coefficient (DSC). GTV/OAR contours were compared with an intraobserver approach and compared to predefined reference structures.

RESULTS

Mean volume for GTV_prostate in the intraobserver approach decreased from 87 ± 44 cm³ (native CT) to 75 ± 22 cm³ (CT-iMAR) (P = 0.168). Compared to the reference, DSC values for GTV_{P rostate} increased from 0.68 ± 0.15 to 0.78 ± 0.07 (CT vs. iMAR) (P < 0.05). In the H & N region, the reference for GTV_{T ongue} (34 cm³ ) was underestimated on both data sets. No significant improvement in DSC values (0.83 ± 0.06 (native CT) versus 0.86 ± 0.06 (CT-iMAR)) was observed.

CONCLUSION

The use of iMAR improves the anatomical delineation at the transition of prostate and bladder in cases of bilateral hip replacement. In the H & N region, anatomical residual structures and experience were apparently sufficient for precise contouring.

T2* Relaxometry in Patients with Parkinson's Disease : Use of an Automated Atlas-based Approach

BACKGROUND

Magnetic resonance (MR) relaxometry is of increasing scientific relevance in neurodegenerative disorders but is still not established in clinical routine. Several studies have investigated relaxation time alterations in disease-specific areas in Parkinson's disease (PD), all using manually drawn regions of interest (ROI). Implementing MR relaxometry into the clinical setting involves the reduction of time needed for postprocessing using an investigator-independent and reliable approach. The aim of this study was to evaluate an automated, atlas-based ROI method for evaluating T2* relaxation times in patients with PD.

METHOD

Automated atlas-based ROI analysis of quantitative T2* maps were generated from 20 PD patients and 20 controls. To test for the accuracy of the atlas-based ROI segmentation, we evaluated the spatial overlap in comparison with manually segmented ROIs using the Dice similarity coefficient (DSC). Additionally, we tested for group differences using our automated atlas-based ROIs of the putamen, globus pallidus, and substantia nigra.

RESULTS

A good spatial overlap accuracy was shown for the automated segmented putamen (mean DSC, 0.64 ± 0.04) and was inferior but still acceptable for the substantia nigra (mean DSC, 0.50 ± 0.17). Based on our automated defined ROI selection, a significant decrease of T2* relaxation time was found in the putamen as well as in the internal and external globus pallidus in PD patients compared with healthy controls.

CONCLUSION

Automated digital brain atlas-based approaches are reliable, more objective and time-efficient, and therefore have the potential to replace the time-consuming manual drawing of ROIs.

Identifying radiotherapy target volumes in brain cancer by image analysis

To establish the optimal radiotherapy fields for treating brain cancer patients, the tumour volume is often outlined on magnetic resonance (MR) images, where the tumour is clearly visible, and mapped onto computerised tomography images used for radiotherapy planning. This process requires considerable clinical experience and is time consuming, which will continue to increase as more complex image sequences are used in this process. Here, the potential of image analysis techniques for automatically identifying the radiation target volume on MR images, and thereby assisting clinicians with this difficult task, was investigated. A gradient-based level set approach was applied on the MR images of five patients with grades II, III and IV malignant cerebral glioma. The relationship between the target volumes produced by image analysis and those produced by a radiation oncologist was also investigated. The contours produced by image analysis were compared with the contours produced by an oncologist and used for treatment. In 93% of cases, the Dice similarity coefficient was found to be between 60 and 80%. This feasibility study demonstrates that image analysis has the potential for automatic outlining in the management of brain cancer patients, however, more testing and validation on a much larger patient cohort is required.

Availability of a simplified lung ventilation imaging algorithm based on four-dimensional computed tomography

PURPOSE

It is still not conclusive which four-dimensional computed tomography (4DCT)-based ventilation imaging algorithm is most accurate and efficient. In this study, we proposed a simplified algorithm (VIAAVG) which only requires the average computed tomography (AVG CT) as input, and quantitatively compared its accuracy and efficiency with three other popular algorithms.

MATERIAL AND METHODS

Fifty patients with lung or esophageal cancer who underwent radiotherapy were enrolled. Single photon emission computed tomography (SPECT) ventilation images (VI-SPECT) and 4DCT were acquired 1-3 days before the first treatment session. The end of exhalation and the end of inhalation CT were registered to derive deformable vector field (DVF) using MIMvista. 4DCT-based ventilation images (CTVI) were first calculated respectively by means of four algorithms (VIAJAC, VIAHU, VIAPRO and VIAAVG). The computation times were compared using paired t-test. The corresponding CTVIs (CTVIJAC, CTVIHU, CTVIPRO and CTVIAVG) and VI-SPECT were segmented into three equal sub-volumes (high, medium and low function lung, respectively) after smoothing and normalization. The Dice Similarity Coefficients (DSCs) were calculated for each sub-volume between each CTVI and VI-SPECT. The average DSCs for high, medium and low function lung in different CTVIs for each patient were compared using paired t-test.

RESULTS

The mean DSCs for CTVIJAC, CTVIHU, CTVIPRO and CTVIAVG were 0.3255, 0.4465, 0.5865 and 0.5958, respectively. The average computation times for CTVIJAC, CTVIHU, CTVIPRO and CTVIAVG were 18.3 s, 24.2 s, 144.8 s and 15.0 s.

CONCLUSION

VIAAVG is available for clinical use because of its high accuracy, improved efficiency and less input requirement compared to the other algorithms.

A registration method for improving quantitative assessment in probabilistic diffusion tractography

Diffusion MRI-based probabilistic tractography is a powerful tool for non-invasively investigating normal brain architecture and alterations in structural connectivity associated with disease states. Both voxelwise and region-of-interest methods of analysis are capable of integrating population differences in tract amplitude (streamline count or density), given proper alignment of the tracts of interest. However, quantification of tract differences (between groups, or longitudinally within individuals) has been hampered by two related features of white matter. First, it is unknown to what extent healthy individuals differ in the precise location of white matter tracts, and to what extent experimental factors influence perceived tract location. Second, white matter lacks the gross neuroanatomical features (e.g., gyri, histological subtyping) that make parcellation of grey matter plausible - determining where tracts "should" lie within larger white matter structures is difficult. Accurately quantifying tractographic connectivity between individuals is thus inherently linked to the difficulty of identifying and aligning precise tract location. Tractography is often utilized to study neurological diseases in which the precise structural and connectivity abnormalities are unknown, underscoring the importance of accounting for individual differences in tract location when evaluating the strength of structural connectivity. We set out to quantify spatial variance in tracts aligned through a standard, whole-brain registration method, and to assess the impact of location mismatch on groupwise assessments of tract amplitude. We then developed a method for tract alignment that enhances the existing standard whole brain registration, and then tested whether this method improved the reliability of groupwise contrasts. Specifically, we conducted seed-based probabilistic diffusion tractography from primary motor, supplementary motor, and visual cortices, projecting through the corpus callosum. Streamline counts decreased rapidly with movement from the tract center (-35% per millimeter); tract misalignment of a few millimeters caused substantial compromise of amplitude comparisons. Alignment of tracts "peak-to-peak" is essential for accurate amplitude comparisons. However, for all transcallosal tracts registered through the whole-brain method, the mean separation distance between an individual subject's tract and the average tract (3.2 mm) precluded accurate comparison: at this separation, tract amplitudes were reduced by 74% from peak value. In contrast, alignment of subcortical tracts (thalamo-putaminal, pallido-rubral) was substantially better than alignment for cortical tracts; whole-brain registration was sufficient for these subcortical tracts. We demonstrated that location mismatches in cortical tractography were sufficient to produce false positive and false negative amplitude estimates in both groupwise and longitudinal comparisons. We then showed that our new tract alignment method substantially reduced location mismatch and improved both reliability and statistical power of subsequent quantitative comparisons.

SCreg: a registration-based platform to compare unicondylar knee arthroplasty SPECT/CT scans

Background

A combination of conventional computed tomography and single photon emitted computed tomography (SPECT/CT) provides simultaneous data on the intensity and location of osteoblastic activity. Currently, since SPECT/CT scans are not spatially aligned, scans following knee arthroplasty are compared by extracting average and maximal values of osteoblastic activity intensity from large subregions of the structure of interest, which leads to a loss of resolution, and hence, information. Therefore, this paper describes the SPECT/CT registration platform (SCreg) based on the principle of image registration to spatially align SPECT/CT scans following unicondylar knee arthroplasty (UKA) and allow full resolution intra-subject and inter-subject comparisons.

Methods

SPECT-CT scans of 20 patients were acquired before and 1 year after UKA. Firstly, scans were pre-processed to account for differences in voxel sizes and divided in volumes of interest. This was followed by optimization of registration parameters according to their volumetric agreement, and alignment using a combination of rigid, affine and non-rigid registration. Finally, radiotracer uptakes were normalized, and differences between pre-operative and post-operative activity were computed for each voxel. Wilcoxon signed rank sum test was performed to compare Dice similarity coefficients pre- and post-registration.

Results

Qualitative and quantitative validation of the platform assessing the correct alignment of SPECT/CT scans resulted in Dice similarity coefficient values over 80% and distances between predefined anatomical landmarks below the fixed threshold of (2;2;0) voxels. Locations of increased and decreased osteoblastic activity obtained during comparisons of osteoblastic activity before and after UKA were mainly consistent with literature.

Conclusions

Thus, a full resolution comparison performed on the platform could assist surgeons and engineers in optimizing surgical parameters in view of bone remodeling, thereby improving UKA survivorship.

Prognostic factors associated with the accuracy of deformable image registration in lung cancer patients treated with stereotactic body radiotherapy

We evaluated the accuracy of an in-house program in lung stereotactic body radiation therapy (SBRT) cancer patients, and explored the prognostic factors associated with the accuracy of deformable image registrations (DIRs). The accuracy of the 3 programs which implement the free-form deformation and the B-spline algorithm was compared regarding the structures on 4-dimensional computed tomography (4DCT) image datasets between the peak-inhale and peak-exhale phases. The dice similarity coefficient (DSC) and normalized DSC (NDSC) were measured for the gross tumor volumes from 19 lung SBRT patients. We evaluated the accuracy of DIR using gross tumor volume, magnitude of displacement from 0% phase to 50% phase, whole lung volume in the 50% phase image, and status of tumor pleural attachment. The median NDSC values using the NiftyReg, MIM Maestro and Velocity AI programs were 1.027, 1.005, and 0.946, respectively, indicating that NiftyReg and MIM Maestro programs had similar accuracy with an uncertainty of < 1 mm. Larger uncertainty of 1 to 2 mm was observed using the Velocity AI program. The NiftyReg and the MIM programs provided higher NDSC values than the median values when the gross tumor volume was attached to the pleura (p <0.05). However, it showed a different trend in using the Velocity AI program. All software programs provided unexpected results, and there is a possibility that such results would reduce the accuracy of 4D treatment planning and adaptive radiotherapy. The unexpected results may be because the tumors are surrounded by other tissues, and there are differences regarding the region of interest for rigid and nonrigid registration. Furthermore, our results indicated that the pleural attachment status might be an important predictor of DIR accuracy for thoracic images, indicating that there is a potentially large dose distribution discrepancy concerning 4D treatment planning and adaptive radiotherapy.

Test-retest reliability of a finger-tapping fMRI task in a healthy population

Measuring brain activity during functional MRI (fMRI) tasks is one of the main tools to identify brain biomarkers of disease or neural substrates associated with specific symptoms. However, identifying correct biomarkers relies on reliable measures. Recently, poor reliability was reported for task-based fMRI measures. The present study aimed to demonstrate the reliability of a finger-tapping fMRI task across two sessions in healthy participants. Thirty-one right-handed healthy participants aged 18-60 years took part in two MRI sessions 3 weeks apart during which we acquired finger-tapping task-fMRI. We examined the overlap of activations between sessions using Dice similarity coefficients, assessing their location and extent. Then, we compared amplitudes calculating intraclass correlation coefficients (ICCs) in three sets of regions of interest (ROIs) in the motor network: literature-based ROIs (10-mm-radius spheres centred on peaks of an activation likelihood estimation), anatomical ROIs (regions as defined in an atlas) and ROIs based on conjunction analyses (superthreshold voxels in both sessions). Finger tapping consistently activated expected regions, for example, left primary sensorimotor cortices, premotor area and right cerebellum. We found good-to-excellent overlap of activations for most contrasts (Dice coefficients: .54-.82). Across time, ICCs showed large variability in all ROI sets (.04-.91). However, ICCs in most ROIs indicated fair-to-good reliability (mean = .52). The least specific contrast consistently yielded the best reliability. Overall, the finger-tapping task showed good spatial overlap and fair reliability of amplitudes on group level. Although caution is warranted in interpreting correlations of activations with other variables, identification of activated regions in response to a task and their between-group comparisons are still valid and important modes of analysis in neuroimaging to find population tendencies and differences.

An investigation into the range dependence of target delineation strategies for stereotactic lung radiotherapy

Background

The “gold standard” approach for defining an internal target volume (ITV) is using 10 gross tumor volume (GTV) phases delineated over the course of one respiratory cycle. However, different sites have adopted several alternative techniques which compress all temporal information into one CT image set to optimize work flow efficiency. The purpose of this study is to evaluate alternative target segmentation strategies with respect to the 10 phase gold standard.

Methods

A Quasar respiratory motion phantom was employed to simulate lung tumor movement. Utilizing 4DCT imaging, a gold standard ITV was created by merging 10 GTV time resolved image sets. Four alternative planed ITV’s were compared using free breathing (FB), average intensity projection (AIP), maximum image projection (MIP), and an augmented FB (FB-Aug) set where the ITV included structures from FB plus max-inhale/exhale image sets. Statistical analysis was performed using the Dice similarity coefficient (DSC). Seventeen patients previously treated for lung SBRT were also included in this retroactive study.

Results

PTV’s derived from the FB image set are the least comparable with the 10 phase benchmark (DSC = 0.740-0.408). For phantom target motion greater than 1 cm, FB and AIP ITV delineation exceeded the 10 phase benchmark by 2% or greater, whereas MIP target segmentation was found to be consistently within 2% agreement with the gold standard (DSC > 0.878). Clinically, however, the FB-Aug method proved to be most favorable for tumor movement up to 2 cm (DSC = 0.881 ± 0.056).

Conclusion

Our results indicate the range of tumor motion dictates the accuracy of the defined PTV with respect to the gold standard. When considering delineation efficiency relative to the 10 phase benchmark, the FB-Aug technique presents a potentially proficient and viable clinical alternative. Among various techniques used for image segmentation, a judicious balance between accuracy and efficiency is inherently required to account for tumor trajectory, range and rate of mobility.

Improving interobserver agreement and performance of deep learning models for segmenting acute ischemic stroke by combining DWI with optimized ADC thresholds

OBJECTIVES

To examine the role of ADC threshold on agreement across observers and deep learning models (DLMs) plus segmentation performance of DLMs for acute ischemic stroke (AIS).

METHODS

Twelve DLMs, which were trained on DWI-ADC-ADC combination from 76 patients with AIS using 6 different ADC thresholds with ground truth manually contoured by 2 observers, were tested by additional 67 patients in the same hospital and another 78 patients in another hospital. Agreement between observers and DLMs were evaluated by Bland-Altman plot and intraclass correlation coefficient (ICC). The similarity between ground truth (GT) defined by observers and between automatic segmentation performed by DLMs was evaluated by Dice similarity coefficient (DSC). Group comparison was performed using the Mann-Whitney U test. The relationship between the DSC and ADC threshold as well as AIS lesion size was evaluated by linear regression analysis. A p < .05 was considered statistically significant.

RESULTS

Excellent interobserver agreement and intraobserver repeatability in the manual segmentation (all ICC > 0.98, p < .001) were achieved. The 95% limit of agreement was reduced from 11.23 cm2 for GT on DWI to 0.59 cm2 for prediction at an ADC threshold of 0.6 × 10-3 mm2/s combined with DWI. The segmentation performance of DLMs was improved with an overall DSC from 0.738 ± 0.214 on DWI to 0.971 ± 0.021 on an ADC threshold of 0.6 × 10-3 mm2/s combined with DWI.

CONCLUSIONS

Combining an ADC threshold of 0.6 × 10-3 mm2/s with DWI reduces interobserver and inter-DLM difference and achieves best segmentation performance of AIS lesions using DLMs.

KEY POINTS

• Higher Dice similarity coefficient (DSC) in predicting acute ischemic stroke lesions was achieved by ADC thresholds combined with DWI than by DWI alone (all p < .05). • DSC had a negative association with the ADC threshold in most sizes, both hospitals, and both observers (most p < .05) and a positive association with the stroke size in all ADC thresholds, both hospitals, and both observers (all p < .001). • An ADC threshold of 0.6 × 10-3 mm2/s eliminated the difference of DSC at any stroke size between observers or between hospitals (p = .07 to > .99).

A validation framework to assess performance of commercial deformable image registration in lung radiotherapy

INTRODUCTION

Deformable image registration (DIR) can play an important role in the context of adaptive radiotherapy. The AAPM Task Group 132 (TG-132) has described several quantitative measures for DIR error assessment but they can only be accurately defined when there is a ground-truth present in high-contrast regions. This work aims to set out a framework to obtain optimal results for CT-CT lung DIR in clinical setting for a commercially available system by quantifying the DIR performance in both low- and high-contrast regions.

METHODS

Five publicly available thorax datasets were used to assess the DIR quality. A "Ghost fiducial" method was implemented by windowing the contrast in a new feature provided by Varian Velocity v4.1. Target registration error (TRE) of the landmarks and Dice-similarity coefficient of the tumour were calculated at three different contrast settings to assess the algorithm in high- and low-contrast scenarios.

RESULTS

For the original unedited dataset, higher resolution DIR methods showed best performance acceptable within the recommended limit according to TG-132, when actual displacements were less than 10 mm. The relation of the actual displacement of the landmarks and TRE shows the limited capacity of the algorithm to deal with movements larger than 10 mm.

CONCLUSION

This work found the performance of DIR methods and settings available in Varian Velocity v4.1 to be a function of contrast level as well as extent of motion. This highlights the need for multiple metrics to assess different aspects of DIR performance for various applications related to low-contrast and/or high-contrast regions.

Dynamic hybrid-phase computed tomography simulation in lung stereotactic body radiotherapy: A feasibility study

To assess the feasibility of dynamic hybrid-phase computed tomography (CT_DHP) simulation when patients undergo lung stereotactic body radiation therapy (SBRT). Eighteen non-small-cell lung-cancer patients were immobilised in a stereotactic body frame with abdominal compression. All underwent dynamic hybrid-phase CT scans that were compared with cone-beam CT (CBCT). We also determined the internal target volume (ITV) and evaluated the following four metrics: the "AND" function in the Boolean module of Eclipse, volume overlap (VO), Dice similarity coefficient (DSC), and dose-volume histogram. The average ITV values of 4DCT_DHP and 3D-CBCT were respectively 12.82±10.42 and 14.6±12.18 cm³ (n=72, p<0.001), and the average ITV value of AND was 11.7±10.1 cm³. The average planning target volume (PTV) of 4DCT_DHP and 3D-CBCT was 25.63±18.04 and 28.00±19.82 cm³ (n=72, p<0.001). The median AND difference between ITV and PTV was significant (p<0.01) and had a significantly linear distribution (R²=0.991 for ITV, R²=0.972 for PTV). The average VO of PTV was greater than that of ITV (0.81±0.096; 0.78±0.11). We also observed that the average DSC in PTV (0.83±0.066) was greater than that in ITV (0.81±0.084). The average results indicated that 97.9%±3.44 of ITV_CBCT was covered by 95% of the prescribed dose. The average minimum, maximum and mean percentage doses of ITV_CBCT were 87.9%±9.46, 107.3%±1.57, and 101.3%±1.12, respectively. This paper has demonstrated that dynamic hybrid-phase CT simulation for patients undergoing lung SBRT and also published evaluation metrics in scientific analysis. Our approach also has the advantage of adequate margin and fewer phases in CT simulation.

Graph deep network for optic disc and optic cup segmentation for glaucoma disease using retinal imaging

The fundus imaging method of eye screening detects eye diseases by segmenting the optic disc (OD) and optic cup (OC). OD and OC are still challenging to segment accurately. This work proposes three-layer graph-based deep architecture with an enhanced fusion method for OD and OC segmentation. CNN encoder-decoder architecture, extended graph network, and approximation via fusion-based rule are explored for connecting local and global information. A graph-based model is developed for combining local and overall knowledge. By extending feature masking, regularization of repetitive features with fusion for combining channels has been done. The performance of the proposed network is evaluated through the analysis of different metric parameters such as dice similarity coefficient (DSC), intersection of union (IOU), accuracy, specificity, sensitivity. Experimental verification of this methodology has been done using the four benchmarks publicly available datasets DRISHTI-GS, RIM-ONE for OD, and OC segmentation. In addition, DRIONS-DB and HRF fundus imaging datasets were analyzed for optimizing the model's performance based on OD segmentation. DSC metric of methodology achieved 0.97 and 0.96 for DRISHTI-GS and RIM-ONE, respectively. Similarly, IOU measures for DRISHTI-GS and RIM-ONE datasets were 0.96 and 0.93, respectively, for OD measurement. For OC segmentation, DSC and IOU were measured as 0.93 and 0.90 respectively for DRISHTI-GS and 0.83 and 0.82 for RIM-ONE data. The proposed technique improved value of metrics with most of the existing methods in terms of DSC and IOU of the results metric of the experiments for OD and OC segmentation.

A deep learning algorithm for automated adrenal gland segmentation on non-contrast CT images

BACKGROUND

The adrenal glands are small retroperitoneal organs, few reference standards exist for adrenal CT measurements in clinical practice. This study aims to develop a deep learning (DL) model for automated adrenal gland segmentation on non-contrast CT images, and to conduct a preliminary large-scale study on age-related volume changes in normal adrenal glands using the model output values.

METHODS

The model was trained and evaluated on a development dataset of annotated non-contrast CT scans of bilateral adrenal glands, utilizing nnU-Net for segmentation task. The ground truth was manually established by two experienced radiologists, and the model performance was assessed using the Dice similarity coefficient (DSC). Additionally, five radiologists provided annotations on a subset of 20 randomly selected cases to measure inter-observer variability. Following validation, the model was applied to a large-scale normal adrenal glands dataset to segment adrenal glands.

RESULTS

The DL model development dataset contained 1301 CT examinations. In the test set, the median DSC scores for the segmentation model of left and right adrenal glands were 0.899 and 0.904 respectively, and in the independent test set were 0.900 and 0.896. Inter-observer DSC for radiologist manual segmentation did not differ from automated machine segmentation (P = 0.541). The large-scale normal adrenal glands dataset contained 2000 CT examinations, the graph shows that adrenal gland volume increases first and then decreases with age.

CONCLUSION

The developed DL model demonstrates accurate adrenal gland segmentation, and enables a comprehensive study of age-related adrenal gland volume variations.

Optical Surface Management System and BladderScan for Patient Setup During Radiotherapy of Postoperative Prostate Cancer

Background: The precision of postoperative prostate cancer radiotherapy is significantly influenced by setup errors and alterations in bladder morphology. Utilizing daily cone beam computed tomography (CBCT) imaging allows for the correction of setup errors. However, this naturally leads to the question of the issue of peripheral dose and workload. Thus, a zero-dose, noninvasive technique to reproduce the bladder volume and improve patient setup accuracy was needed. Purpose: The aim of this study is to investigate if the setup method by combining Optical Surface Management System (OSMS) and BladderScan can improve the accuracy of setup and accurately reproduce the bladder volume during radiotherapy of postoperative prostate cancer and to guide CTV-PTV margins for clinic. Method: The experimental group consisted of 15 postoperative prostate cancer patients who utilized a setup method that combined OSMS and BladderScan. This group recorded 103 setup errors, verified by CBCT. The control group comprised 25 patients, among whom 114 setup errors were recorded using the conventional setup method involving skin markers; additionally, patients in this group also exhibited spontaneous urinary suppression. The errors including lateral (Lat), longitudinal (Lng), vertical directions (Vrt), Pitch, Yaw, and Roll were analyzed between the two methods. The Dice similarity coefficient (DSC) and volume differences of the bladder between CBCT and planning CT were compared as the bladder concordance indicators. Results: The errors in the experimental group at Vrt, Lat, and Lng were 0.17 ± 0.12, 0.22 ± 0.17, and 0.18 ± 0.12 cm, and the control group were 0.25 ± 0.15, 0.31 ± 0.21, 0.34 ± 0.22 cm. The rotation errors of Pitch, Roll, and Yaw in the experimental group were 0.18 ± 0.12°, 0.11 ± 0.1°, and 0.18 ± 0.13°, and in the control group, they were 0.96 ± 0.89°, 1.01 ± 0.86°, and 1.02 ± 0.84°. The DSC and volume differences were 92.52 ± 1.65% and 39.99 ± 28.75 cm3 in the patients with BladderScan, and in the control group, they were 62.98 ± 22.33%, 273.89 ± 190.62 cm3. The P < 0.01 of the above performance indicators indicates that the difference is statistically significant. Conclusion: The accuracy of the setup method by combining OSMS and BladderScan was validated by CBCT in our study. The method in our study can improve the setup accuracy during radiotherapy of postoperative prostate cancer compared to the conventional setup method.

Variation assessment of deformable registration in stereotactic radiosurgery

INTRODUCTION

The regular functions of CT-MRI registration include delineation of targets and organs-at-risk (OARs) in radiosurgery planning. The question of whether deformable image registration (DIR) could be applied to stereotactic radiosurgery (SRS) in its place remains a subject of debate.

METHODS

This study collected data regarding 16 patients who had undergone single-fraction SRS treatment. All lesions were located close to the brainstem. CT and MRI two image sets were registered by both rigid image registration (RIR) and DIR algorithms. The contours of the OARs were drawn individually on the rigid and deformable CT-MRI image sets by qualified radiation oncologists and dosimetrists. The evaluation metrics included volume overlapping (VO), Dice similarity coefficient (DSC), and dose. The modified demons deformable algorithm (VARIAN SmartAdapt) was used for evaluation in this study.

RESULTS

The mean range of VO for OARs was 0.84 ± 0.08, and DSC was 0.82 ± 0.07. The maximum average volume difference was at normal brain (17.18 ± 14.48 cm³) and the second highest was at brainstem (2.26 cm³ ± 1.18). Pearson correlation testing showed that all DIRs' OAR volumes were linearly and significantly correlated with RIRs' volume (0.679-0.992, two tailed, P << 0.001). The 100% dose was prescribed at gross tumor volume (GTV). The average maximum percent dose difference was observed in brainstem (26.54% ± 27.027), and the average mean dose difference has found at same organ (1.6% ± 1.66).

CONCLUSION

The change in image-registration method definitely produces dose variance, and is significantly more what depending on the target location. The volume size of OARs, however, was not statistical significantly correlated with dose variance.

[Feasibility of Evaluating Result of Auto-segmentation of Target Volumes in Radiotherapy with Medical Consideration Index]

OBJECTIVE

To explore the feasibility of using the bidirectional local distance based medical similarity index (MSI) to evaluate automatic segmentation on medical images.

METHODS

Taking the intermediate risk clinical target volume for nasopharyngeal carcinoma manually segmented by an experience radiation oncologist as region of interest, using Atlas-based and deep-learning-based methods to obtain automatic segmentation respectively, and calculated multiple MSI and Dice similarity coefficient (DSC) between manual segmentation and automatic segmentation. Then the difference between MSI and DSC was comparatively analyzed.

RESULTS

DSC values for Atlas-based and deep-learning-based automatic segmentation were 0.73 and 0.84 respectively. MSI values for them varied between 0.29~0.78 and 0.44~0.91 under different inside-outside-level.

CONCLUSIONS

It is feasible to use MSI to evaluate the results of automatic segmentation. By setting the penalty coefficient, it can reflect phenomena such as under-delineation and over-delineation, and improve the sensitivity of medical image contour similarity evaluation.

Understanding the heterogenicity of unicoronal synostosis - A morphometric analysis of cases compared to controls

BACKGROUND

Preoperative severity of unicoronal synostosis varies greatly and involves the frontal bone, skull base and orbits. Degree of deformity affects long-term morphological and functional outcomes after surgery. The aim of this study was to describe the morphological heterogenicity and investigate its relation to patient-specific factors.

MATERIALS AND METHODS

In this retrospective cohort study, non-syndromic unicoronal synostosis patients treated between 2006 and 2022 at Necker Hospital, France or Uppsala University Hospital, Sweden, were included and matched to controls. Severity of skull base, orbital and posterior skull asymmetry, degree of anterior plagiocephaly and Harlequin deformity, lateralisation, head circumference, age, timing of metopic fusion and fusion of peri-pterionic sutures were investigated.

RESULTS

Ninety-five patients and ninety-three controls were included. Skull base asymmetry was linearly related to orbital asymmetry (p < 0.001), correlated with earlier CT scans (p = 0.004) and anterior (p < 0.001) and posterior (p = 0.03) plagiocephaly. Posterior plagiocephaly was more common in patients (31%) compared with controls (5%) (p < 0.001). A patent metopic suture above nine months of age was associated with severe Harlequin deformity (p = 0.04) and a lower head circumference when fused (p = 0.03). Fronto-sphenoidal suture fusion was associated with later CT scans (p < 0.001) and less skull base asymmetry (p = 0.002). Spheno-parietal fusion was correlated with decreased skull base asymmetry (p = 0.03). Right lateralisation was more common in females.

CONCLUSIONS

Heterogenicity of unicoronal synostosis seems to be predominantly explained by variability in skull base morphology. Peri-pterionic fusions might limit deformity.

Comparison of an in-house hybrid DIR method to NiftyReg on CBCT and CT images for head and neck cancer

An in-house hybrid deformable image registration (DIR) method, which combines free-form deformation (FFD) and the viscous fluid registration method, is proposed. Its results on the planning computed tomography (CT) and the day 1 treatment cone-beam CT (CBCT) image from 68 head and neck cancer patients are compared with the results of NiftyReg, which uses B-spline FFD alone. Several similarity metrics, the target registration error (TRE) of annotated points, as well as the Dice similarity coefficient (DSC) and Hausdorff distance (HD) of the propagated organs at risk are employed to analyze their registration accuracy. According to quantitative analysis on mutual information, normalized cross-correlation, and the absolute pixel value differences, the results of the proposed DIR are more similar to the CBCT images than the NiftyReg results. Smaller TRE of the annotated points is observed in the proposed method, and the overall mean TRE for the proposed method and NiftyReg was 2.34 and 2.98 mm, respectively (p < 0.001). The mean DSC in the larynx, spinal cord, oral cavity, mandible, and parotid given by the proposed method ranged from 0.78 to 0.91, significantly higher than the NiftyReg results (ranging from 0.77 to 0.90), and the HD was significantly lower compared to NiftyReg. Furthermore, the proposed method did not suffer from unrealistic deformations as the NiftyReg did in the visual evaluation. Meanwhile, the execution time of the proposed method was much higher than NiftyReg (96.98 ± 11.88 s vs. 4.60 ± 0.49 s). In conclusion, the in-house hybrid method gave better accuracy and more stable performance than NiftyReg.