Dice Similarity Coefficient Values Research Articles

Clinical application and improvement of a CNN-based autosegmentation model for clinical target volumes in cervical cancer radiotherapy.

ObjectiveClinical target volume (CTV) autosegmentation for cervical cancer is desirable for radiation therapy. Data heterogeneity and interobserver variability (IOV) limit the clinical adaptability of such methods. The adaptive method is proposed to improve the adaptability of CNN‐based autosegmentation of CTV contours in cervical cancer.MethodsThis study included 400 cervical cancer treatment planning cases with CTV delineated by radiation oncologists from three hospitals. The datasets were divided into five subdatasets (80 cases each). The cases in datasets 1, 2, and 3 were delineated by physicians A, B, and C, respectively. The cases in datasets 4 and 5 were delineated by multiple physicians. Dataset 1 was divided into training (50 cases), validation (10 cases), and testing (20 cases) cohorts, and they were used to construct the pretrained model. Datasets 2–5 were regarded as host datasets to evaluate the accuracy of the pretrained model. In the adaptive process, the pretrained model was fine‐tuned to measure improvements by gradually adding more training cases selected from the host datasets. The accuracy of the autosegmentation model on each host dataset was evaluated using the corresponding test cases. The Dice similarity coefficient (DSC) and 95% Hausdorff distance (HD_95) were used to evaluate the accuracy.ResultsBefore and after adaptive improvements, the average DSC values on the host datasets were 0.818 versus 0.882, 0.763 versus 0.810, 0.727 versus 0.772, and 0.679 versus 0.789, which are improvements of 7.82%, 6.16%, 6.19%, and 16.05%, respectively. The average HD_95 values were 11.143 mm versus 6.853 mm, 22.402 mm versus 14.076 mm, 28.145 mm versus 16.437 mm, and 33.034 mm versus 16.441 mm, which are improvements of 37.94%, 37.17%, 41.60%, and 50.23%, respectively.ConclusionThe proposed method improved the adaptability of the CNN‐based autosegmentation model when applied to host datasets.

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

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

Reproducibility of manual segmentation in muscle imaging.

PurposeTo assess the reproducibility of a manual muscle MRI segmentation method that follows a specific set of recommendations developed in our center.Materials and methodsNine healthy volunteers underwent a muscle MRI examination that included a TSE T2 sequence of the thighs. Muscle segmentation was performed by three operators: an expert operator (OP1) with 3 years of experience and two radiology residents (OP2 and 3) who were both given basic segmentation instructions, whereas only OP2 underwent additional supervised training from OP1. Intra- and inter-operator Dice similarity coefficient (DSC) was calculated.ResultsOP1 showed the highest average intra-operator DSC values (0.885), whereas OP2 had higher average DSC (0.856) compared to OP3 (0.818). The highest inter-operator agreement was observed between Operators 1 and 2 (0.814) and the lowest between OP2 and OP3 (0.702). Confidence interval (CI) analysis showed that the most experienced operator also had the least variability in drawing the ROIs, whereas OP2 showed both higher intra-operator reproducibility compared to OP3 and higher inter-operator agreement with OP1. The muscles that showed the least reproducibility were the semimembranosus and the short head of the biceps femoris.DiscussionFollowing specific recommendations such as these ones derived from our single-center experience leads to an overall high reproducibility of manual muscle segmentation and is helpful in improving both intra-operator and inter-operator reproducibility in less experienced operators.

Feasibility of magnetic resonance-only high-dose-rate surface brachytherapy for clinical application.

In this article, we investigate the feasibility of magnetic resonance (MR)-only imaging for high-dose-rate (HDR) surface brachytherapy (SABT). We examined whether a standard CT-based planning can be replaced with an MR-only planning. For this purpose, the MRI digitization and plan quality check processes were compared against the standard CT-based processes. A prospective clinical implementation of the MR-only planning was evaluated on a clinical data set. A pointwise encoding time reduction with radial acquisition (PETRA) sequence was optimized for visualization of Freiburg flap (FF) on MR images. MR and conventional CT images were acquired with a FF applicator (Elekta, Stockholm, Sweden) placed on the following phantoms: (1) flat styrofoam (FST), FF locked-in placed with supporting structure; (2) cast-made facemask, and (3) porcine leg (PL). Catheters were digitized and activated with 10mm step size on Oncentra Brachy 4.5.3 Treatment Planning System. The CT-only and MR-only treatment plans were generated by optimizing the dose to the target defined as volume at 3mm skin depth. To compare the plans, the MRI-to-CT alignment was performed via rigid registration. Positional displacements of dwell positions between CT and MR plans were compared on the FST phantom and the relative percent dose difference in 2210 different points from CT or MR-only plans was compared. For all three phantoms, the comparabilities between CT and MR-only plans were assessed by calculating dice similarity coefficient (DSC) for volumes enclosing 150%, 125%, 100%, 95%, 90%, 80%, and 65% isodose lines (V150 -V65 ). The MR images of FF placed on the forearm of a healthy subject were acquired with this optimized PETRA sequence and used for treatment planning. The relative percent dose was calculated on 140 representative points placed at 3mm skin depth to evaluate the dose to the skin. Using the optimized PETRA sequence, MRTP digitization accuracy was<1mm in each dimension and on three-dimensional (3D) displacement for the FST phantom. In each phantom and clinical data set, it was possible to generate MR-only treatment plans with the 3mm skin depth prescription. In the FST phantom, the mean relative dose at the points was not significantly different(<0.1% difference) for CT or MR-based plans. The assessment of similarities in dose profiles between CT and MR-only plans' provided DSC values greater than 0.96, 0.92, and 0.73 for all volumes enclosing up to 100%, 125%, and 150% isodose lines, respectively. The feasibility of generating a HDR treatment plan with FF using MR-only has been evaluated in phantoms with varying geometry and for a clinical data set. The optimization of a standard MRI sequence-PETRA-implemented in this study showed that FF-based catheters can be digitized and a plan can be generated using only MRI. The resulting MR-only plans were comparable to the conventional CT-based plans, suggesting that MRI alone can generate clinically acceptable plans for FF in phantoms and on a clinical data set. Reliable MR-only treatment planning could improve treatment prescription through more accurate characterization of soft tissue targets.

Image Features of Magnetic Resonance Angiography under Deep Learning in Exploring the Effect of Comprehensive Rehabilitation Nursing on the Neurological Function Recovery of Patients with Acute Stroke.

This study was to explore the effects of imaging characteristics of magnetic resonance angiography (MRA) based on deep learning on the comprehensive rehabilitation nursing on the neurological recovery of patients with acute stroke. In this study, 84 patients with acute stroke who were treated in hospital were selected as the research objects, and they were rolled into a control group (routine care) and an experimental group (comprehensive rehabilitation care). The dense dilated block-convolution neural network (DD-CNN) algorithm under deep learning for cerebrovascular was adopted to assess the effect of comprehensive rehabilitation care on the neurological recovery of patients with acute stroke. The results showed that the Berg scale scores, Fugl-Meyer scores, and Functional Independence Measure (FIM) scores of the experimental group of patients after 6 weeks and 12 weeks of comprehensive rehabilitation nursing were greatly different from those before treatment, showing statistical differences (P < 0.05). Compared with conventional magnetic resonance imaging (MRI) images, MRA images based on CNN algorithm, Dense Net algorithm, and DD-CNN algorithm can more clearly show the patient's cerebral artery occlusion. The average dice similarity coefficient (DSC) values of CNN algorithm, Dense Net algorithm, and DD-CNN algorithm were determined to be 84.3%, 95.7%, and 97.8%, respectively; the average sensitivity (Sen) values of the three algorithms were 76.1%, 95.4%, and 96.8%, respectively; and the average accuracy (Acc) values were 87.9%, 96.3%, and 97.9%, respectively. Thus, there were statistically obvious differences among the three algorithms in terms of average values of DSC, Sen, and Acc (P < 0.05). The MRA images processed by the DD-CNN algorithm showed that the degree of neurological recovery of the experimental group was observably greater than that of the control group, and the difference was statistically obvious (P < 0.05). In short, the image features of MRA based on the deep learning DD-CNN algorithm showed good application value in studying the effect of comprehensive rehabilitation nursing on the neurological recovery of patients with acute stroke, and it was worthy of promotion.

Deep generative models for automated muscle segmentation in computed tomography scanning.

Accurate gluteus medius (GMd) volume evaluation may aid in the analysis of muscular atrophy states and help gain an improved understanding of patient recovery via rehabilitation. However, the segmentation of muscle regions in GMd images for cubic muscle volume assessment is time-consuming and labor-intensive. This study automated GMd-region segmentation from the computed tomography (CT) images of patients diagnosed with hip osteoarthritis using deep learning and evaluated the segmentation accuracy. To this end, 5250 augmented pairs of training data were obtained from five participants, and a conditional generative adversarial network was used to identify the relationships between the image pairs. Using the preserved test datasets, the results of automatic segmentation with the trained deep learning model were compared to those of manual segmentation in terms of the dice similarity coefficient (DSC), volume similarity (VS), and shape similarity (MS). As observed, the average DSC values for automatic and manual segmentations were 0.748 and 0.812, respectively, with a significant difference (p < 0.0001); the average VS values were 0.247 and 0.203, respectively, with no significant difference (p = 0.069); and the average MS values were 1.394 and 1.156, respectively, with no significant difference (p = 0.308). The GMd volumes obtained by automatic and manual segmentation were 246.2 cm3 and 282.9 cm3, respectively. The noninferiority of the DSC obtained by automatic segmentation was verified against that obtained by manual segmentation. Accordingly, the proposed GAN-based automatic GMd-segmentation technique is confirmed to be noninferior to manual segmentation. Therefore, the findings of this research confirm that the proposed method not only reduces time and effort but also facilitates accurate assessment of the cubic muscle volume.

International Multi-institutional Patterns of Contouring Practice and Clinical Target Volume Recommendations for Stereotactic Body Radiation Therapy for Non-Spine Bone Metastases

Despite the increasing use of stereotactic body radiation therapy for non-spine bone metastases (NSBM), there is no established standard for target delineation. The objective of this study was to provide consensus recommendations on clinical target volume (CTV) delineation based on international expert contours. Eleven cases of NSBM were contoured by 9 international radiation oncologists. For each case, the gross tumor volume was provided on the simulation computed tomography scans with accompanying magnetic resonance imaging. Participants contoured the CTV and completed a clinical survey. Agreement between CTV contours were analyzed with simultaneous truth and performance level estimation using the kappa coefficient and the Dice similarity coefficient (DSC) and summarized to establish contouring recommendations. A direction-dependent analysis was applied to the consensus contours to quantify margins. All CTV contours were completed. Six participants used a single-dose level, whereas 3 used a 2-dose level simultaneous integrated boost (SIB) technique. For the SIB cases, the largest volume receiving a stereotactic body radiation therapy (SBRT) dose was used for contour analysis. There was substantial agreement between contours across cases with a mean kappa of 0.72 (mean sensitivity 0.85, mean specificity 0.97). The mean DSC value was 0.77 (range, 0.67-0.87). Consensus CTV contouring recommendations were (1) an intraosseous CTV margin of 5 to 10 mm should be strongly considered within contiguous bone; (2) an extraosseous margin of 5 to 10 mm should be strongly considered where there is soft tissue disease or cortical bone disruption; (3) CTVs should be manually cropped to respect anatomic barriers to spread (eg, peritoneal cavity, pleura, uninvolved joint space and cortical bone). CTV contouring recommendations for NSBM-SBRT were established based on analysis of international expert consensus contours with a high level of agreement. These principles may provide guidance to treating physicians and inform future study until prospective clinical data can provide further refinement.

A Comparison of Cranial Cavity Extraction Tools for Non-contrast Enhanced CT Scans in Acute Stroke Patients

Cranial cavity extraction is often the first step in quantitative neuroimaging analyses. However, few automated, validated extraction tools have been developed for non-contrast enhanced CT scans (NECT). The purpose of this study was to compare and contrast freely available tools in an unseen dataset of real-world clinical NECT head scans in order to assess the performance and generalisability of these tools. This study included data from a demographically representative sample of 428 patients who had completed NECT scans following hospitalisation for stroke. In a subset of the scans (n = 20), the intracranial spaces were segmented using automated tools and compared to the gold standard of manual delineation to calculate accuracy, precision, recall, and dice similarity coefficient (DSC) values. Further, three readers independently performed regional visual comparisons of the quality of the results in a larger dataset (n = 428). Three tools were found; one of these had unreliable performance so subsequent evaluation was discontinued. The remaining tools included one that was adapted from the FMRIB software library (fBET) and a convolutional neural network- based tool (rBET). Quantitative comparison showed comparable accuracy, precision, recall and DSC values (fBET: 0.984 ± 0.002; rBET: 0.984 ± 0.003; p = 0.99) between the tools; however, intracranial volume was overestimated. Visual comparisons identified characteristic regional differences in the resulting cranial cavity segmentations. Overall fBET had highest visual quality ratings and was preferred by the readers in the majority of subject results (84%). However, both tools produced high quality extractions of the intracranial space and our findings should improve confidence in these automated CT tools. Pre- and post-processing techniques may further improve these results.

Deep-Learning-Based Automatic Segmentation of Head and Neck Organs for Radiation Therapy in Dogs.

Purpose: This study was conducted to develop a deep learning-based automatic segmentation (DLBAS) model of head and neck organs for radiotherapy (RT) in dogs, and to evaluate the feasibility for delineating the RT planning.Materials and Methods: The segmentation indicated that there were potentially 15 organs at risk (OARs) in the head and neck of dogs. Post-contrast computed tomography (CT) was performed in 90 dogs. The training and validation sets comprised 80 CT data sets, including 20 test sets. The accuracy of the segmentation was assessed using both the Dice similarity coefficient (DSC) and the Hausdorff distance (HD), and by referencing the expert contours as the ground truth. An additional 10 clinical test sets with relatively large displacement or deformation of organs were selected for verification in cancer patients. To evaluate the applicability in cancer patients, and the impact of expert intervention, three methods–HA, DLBAS, and the readjustment of the predicted data obtained via the DLBAS of the clinical test sets (HA_DLBAS)–were compared.Results: The DLBAS model (in the 20 test sets) showed reliable DSC and HD values; it also had a short contouring time of ~3 s. The average (mean ± standard deviation) DSC (0.83 ± 0.04) and HD (2.71 ± 1.01 mm) values were similar to those of previous human studies. The DLBAS was highly accurate and had no large displacement of head and neck organs. However, the DLBAS in the 10 clinical test sets showed lower DSC (0.78 ± 0.11) and higher HD (4.30 ± 3.69 mm) values than those of the test sets. The HA_DLBAS was comparable to both the HA (DSC: 0.85 ± 0.06 and HD: 2.74 ± 1.18 mm) and DLBAS presented better comparison metrics and decreased statistical deviations (DSC: 0.94 ± 0.03 and HD: 2.30 ± 0.41 mm). In addition, the contouring time of HA_DLBAS (30 min) was less than that of HA (80 min).Conclusion: In conclusion, HA_DLBAS method and the proposed DLBAS was highly consistent and robust in its performance. Thus, DLBAS has great potential as a single or supportive tool to the key process in RT planning.

Diffusion Kurtosis-Based Brain Image Segmentation for the Structural Remodeling of White Matter in Patients with Hypoxic-Ischemic Encephalopathy

The aim was to explore the application value of brain image segmentation algorithm based on diffusion kurtosis imaging (DKI) in the structural remodeling of white matter (WM) in patients with hypoxic-ischemic encephalopathy (HIE). 120 patients with leukoencephalopathy and hypoxic-ischemic encephalopathy were selected as the research objects. Their heads were scanned by conventional magnetic resonance imaging (MRI) and DKI sequence. Besides, DKI based on the image segmentation algorithm was applied to process DKI images, DKE software was employed to obtain the values of fractional anisotropy (FA) and mean kurtosis (MK), and the differences in FA and MK were compared between acute and chronic phases. The results showed that the proposed algorithm could realize the best values of Jaccard similarity (JS) and Dice similarity coefficient (DSC) in WM, gray matter (GM), and cerebrospinal fluid (CSF), and the segmentation accuracy was better than other algorithms. FA values at the acute and chronic phases were compared in the area around lesions (0.421 ± 0.065 vs. 0.454 ± 0.052), the posterior limb of internal capsule on the affected side (0.498 ± 0.027 vs. 0.504 ± 0.046), and the pedunculus cerebri on the affected side (0.558 ± 0.038 vs. 0.568 ± 0.042), and the differences were statistically substantial ( P &lt; 0.05 ). Moreover, MK values at the acute and chronic phases were also compared in the area around lesions (1.362 ± 0.098 vs. 1.407 ± 0.077), the centrum semiovale on the affected side (1.305 ± 0.102 vs. 1.343 ± 0.076), the posterior limb of internal capsule on the affected side (1.338 ± 0.543 vs. 1.382 ± 0.076), and the pedunculus cerebri on the affected side (1.329 ± 0.089 vs. 1.398 ± 0.099), showing a statistical meaning ( P &lt; 0.05 ). The results indicated that the changes of FA and MK were related to the structural remodeling of WM. The DKI image segmentation algorithm could be applied in the diagnosis of leukoencephalopathy in patients with hypoxic-ischemic encephalopathy, and DKI technology was of great significance for the research of structural remodeling of WM.

Multimodality-based knee joint modelling method with bone and cartilage structures for total knee arthroplasty.

We propose a robust and accurate knee joint modelling method with bone and cartilage structures to enable accurate surgical guidance for knee surgery. A multimodality registration strategy is proposed to fuse magnetic resonance (MR) and computed tomography (CT) images of the femur and tibia separately to remove spatial inconsistency caused by knee bending in CT/MR scans. Automatic segmentation of the femur, tibia and cartilages is carried out with region of interest clustering and intensity analysis based on the multimodal fusion of images. Experimental results show that the registration error is 1.13±0.30mm. The Dice similarity coefficient values of the proposed segmentation method of the femur, tibia, femoral and tibial cartilages are 0.969, 0.966, 0.910 and 0.872, respectively. This study demonstrates the feasibility and effectiveness of multimodality-based registration and segmentation methods for knee joint modelling. The proposed method can provide users with 3D anatomical models of the femur, tibia, and cartilages with few human inputs.

Postsurgical geometrical variations of tumor bed and brainstem during photon and proton therapy for pediatric tumors of the posterior fossa: dosimetric impact and predictive factors.

Brainstem radionecrosis is an important issue during the irradiation of tumors of the posterior fossa. The aim of the present study is to analyze postsurgical geometrical variations of tumor bed (TB) and brainstem (BS) and their impact on dosimetry. Retrospective collection of data from pediatric patients treated at asingle institution. Availability of presurgical magnetic resonance imaging (MRI) was verified; availability of at least two postsurgical MRIs was considered afurther inclusion criterion. The following metrics were analyzed: total volume, Dice similarity coefficient (DSC), and Haudsdorff distances (HD). Fourteen patients were available for the quantification of major postsurgical geometrical variations of TB. DSC, HD max, and HD average values were 0.47 (range: 0.08;0.76), 11.3mm (7.7;24.5), and 2.6mm (0.7;6.7) between the first and the second postoperative MRI, respectively. Postsurgical geometrical variations of the BS were also observed. Coverage to the TB was reduced in one patient (D95: -2.9 Gy), while D2 to the BS was increased for the majority of patients. Overall, predictive factors for significant geometrical changes were presurgical gross tumor volume (GTV) > 33 mL, hydrocephaly at diagnosis, Luschka foramen involvement, and younger age (≤ 8years). Major volume changes were observed in this cohort, with some dosimetric impact. The use of arecent co-registration MRI is advised. The 2-3 mm HD average observed should be considered in the planning target volume/planning organ at risk volume (PTV/PRV) margin and/or robust optimization planning. Results from wider efforts are needed to verify our findings.

Recognition and Volume Measurement of Intracranial Hematoma through CT Images under Intelligent Recognition Algorithm

In order to realize the automatic recognition of intracranial hematoma and accurate measurement of hematoma volume in patients with ICH (intracerebral hemorrhage), the FCM algorithm (fuzzy c-means algorithm) was improved in this study, and a new level set segmentation algorithm based on FCM was obtained, FCRLS (fuzzy c-means regularized level set). Then, 120 ICH patients were used as the research objects, and the FCRLS algorithm was evaluated by the recall, precise, and F1-score values to evaluate the effect of intracranial hematoma recognition. The CT images of 48 patients with intracranial hematoma were used as the data set of the FCRLS algorithm. The hematoma was segmented, and the DSC (Dice similarity coefficient) value and running time were used to evaluate the segmentation results of the algorithm. At the same time, the LS (level set) algorithm and the FCM algorithm were introduced for comparison. The results show that the recall value of the FCRLS algorithm is 0.89, the precise value is 0.94, the F1-score value is 0.91, the Dice coefficient is 94.81%, and the running time is 14.48 s. Compared with the LS algorithm and the FCM algorithm, the above five indicators have significant differences ( P &lt; 0.05 ). Hematoma volume measurement found that the average error of FCRLS algorithm from expert measurement results was 5.62%, which was statistically significant compared with LS algorithm and FCM algorithm ( P &lt; 0.05 ). In summary, the FCRLS algorithm can accurately identify the cerebral hematoma area of ICH patients, has an ideal segmentation effect on the hematoma, and can accurately measure the true volume of the hematoma, which is worthy of clinical application.

Neural Network Algorithm MRI Images for Analysis of Influencing Factors for Patellar Dislocation in Exercise

The study focused on segmentation effects of the improved algorithm of traditional neural network algorithm, small kernels two-path convolutional neural network (SK-TPCNN) combined with random forest (RF) algorithm on MRI images for patella, and the influencing factors of patellar dislocation during exercise. In this article, the MRI images for patellar dislocation patients were detected by virtue of the neural network algorithm, to establish the patella-related MRI image segmentation algorithm. In terms of dice similarity coefficient (DSC), positive predictive value (PPV), and sensitivity, the detection accuracy of MRI images for patella was evaluated, and the segmentation effect of MRI images for patella was assessed. 30 patients, who were diagnosed as patellar dislocation patients in hospital, were chosen as the research subjects. No matter whether the MRI images of the patients went through the processing of the neural network algorithm or not, all of them were analyzed. The results showed that, among the traditional neural network algorithm, SK-TPCNN algorithm, and SK-TPCNN + RF algorithm, the DSC values were 0.82, 0.71, and 0.79, respectively; the PPV values were 0.77, 0.59, and 0.85, respectively; and the sensitivity values were 0.79, 0.62, and 0.89, respectively. Obviously, the various parameters of the SK-TPCNN + RF algorithm were significantly greater than those of the SK-TPCNN algorithm, and the difference was statistically significant ( P &lt; 0.05 ). It indicated that the segmentation ability of MRI images for patella of the NN algorithm was clearly improved, and the MRI image resolution was dramatically raised, which provided a referable basis for the MRI diagnosis of patients with patellar dislocation during exercise.

Automatic segmentation of uterine endometrial cancer on multi-sequence MRI using a convolutional neural network

Endometrial cancer (EC) is the most common gynecological tumor in developed countries, and preoperative risk stratification is essential for personalized medicine. There have been several radiomics studies for noninvasive risk stratification of EC using MRI. Although tumor segmentation is usually necessary for these studies, manual segmentation is not only labor-intensive but may also be subjective. Therefore, our study aimed to perform the automatic segmentation of EC on MRI with a convolutional neural network. The effect of the input image sequence and batch size on the segmentation performance was also investigated. Of 200 patients with EC, 180 patients were used for training the modified U-net model; 20 patients for testing the segmentation performance and the robustness of automatically extracted radiomics features. Using multi-sequence images and larger batch size was effective for improving segmentation accuracy. The mean Dice similarity coefficient, sensitivity, and positive predictive value of our model for the test set were 0.806, 0.816, and 0.834, respectively. The robustness of automatically extracted first-order and shape-based features was high (median ICC = 0.86 and 0.96, respectively). Other high-order features presented moderate-high robustness (median ICC = 0.57–0.93). Our model could automatically segment EC on MRI and extract radiomics features with high reliability.

Evaluation of Hybrid PET/MRI for Gross Tumor Volume (GTV) Delineation in Colorectal Cancer Liver Metastases Radiotherapy.

PurposeHybrid PET/MRI has been increasingly incorporated into the practice of radiation oncologists since it contains both anatomical and biological data and may bring about personalized radiation plans for each patient. The objective of this study was to evaluate the feasibility of GTV delineation from hybrid PET/MRI compared with that from current-practice MRI during radiotherapy planning in patients with colorectal liver metastases.Patients and MethodsTwenty-four patients (thirty lesions) with colorectal liver metastases were prospectively enrolled in this study. Three physicians delineated the target volume with the most popular delineating methods—the visual method. First of all, differences among the three observers were assessed. The difference and correlation of GTV values obtained by MRI, PET, and hybrid PET/MRI were subjected to statistical analysis afterwards. Finally, the dice similarity coefficient (DSC) was calculated to assess the spatial overlap. Based on the value of DSC, we also evaluate the correlation between DSC and tumor size. GTV-MRI was set as a reference.ResultsThere was no significant difference among observers in GTV-MRI (F=0.118, p=0.889), GTV-PET (F=0.070, p=0.933) and GTV-PET/MRI (F=0.40, p=0.961). 83.33% of GTV-PET/MRI and 63.33% of GTV-PET were larger than the reference GTV-MRI. Statistical analysis revealed that GTV-PET/MRI (p<0.001) and GTV-PET (p<0.05) diverged statistically significantly from GTV-MRI. GTV-PET (r=0.992, p<0.001) and GTV-PET/MRI (r=0.997, p<0.001) were significantly related to GTV-MRI. The average DSC value between GTV-MRI and GTV-PET was 0.51 (range 0–0.90) and that between GTV-MRI and GTV-PET/MRI was 0.72 (range 0.42–0.90). There was a positive correlation between the DSC and GTV-MRI (r=0.851, p<0.05).ConclusionWith the database used, there is good agreement among observers. Hybrid PET/MRI in colorectal liver metastases radiotherapy may affect the GTV delineation. Moreover, the overlap degree between GTV-MRI and GTV-PET/MRI is higher and increases with volume.

Development of a physical geometric phantom for deformable image registration credentialing of radiotherapy centers for a clinical trial.

PurposeThis study aimed to develop a physical geometric phantom for the deformable image registration (DIR) credentialing of radiotherapy centers for a clinical trial and tested the feasibility of the proposed phantom at multiple domestic and international institutions.Methods and materialsThe phantom reproduced tumor shrinkage, rectum shape change, and body shrinkage using several physical phantoms with custom inserts. We tested the feasibility of the proposed phantom using 5 DIR patterns at 17 domestic and 2 international institutions (21 datasets). Eight institutions used the MIM software (MIM Software Inc, Cleveland, OH); seven used Velocity (Varian Medical Systems, Palo Alto, CA), and six used RayStation (RaySearch Laboratories, Stockholm, Sweden). The DIR accuracy was evaluated using the Dice similarity coefficient (DSC) and Hausdorff distance (HD).ResultsThe mean and one standard deviation (SD) values (range) of DSC were 0.909 ± 0.088 (0.434–0.984) and 0.909 ± 0.048 (0.726–0.972) for tumor and rectum proxies, respectively. The mean and one SD values (range) of the HD value were 5.02 ± 3.32 (1.53–20.35) and 5.79 ± 3.47 (1.22–21.48) (mm) for the tumor and rectum proxies, respectively. In three patterns evaluating the DIR accuracy within the entire phantom, 61.9% of the data had more than a DSC of 0.8 in both tumor and rectum proxies. In two patterns evaluating the DIR accuracy by focusing on tumor and rectum proxies, all data had more than a DSC of 0.8 in both tumor and rectum proxies.ConclusionsThe wide range of DIR performance highlights the importance of optimizing the DIR process. Thus, the proposed method has considerable potential as an evaluation tool for DIR credentialing and quality assurance.

A novel specific grading standard study of auto-segmentation of organs at risk in thorax: subjective\u2013objective-combined grading standard

BackgroundTo develop a novel subjective–objective-combined (SOC) grading standard for auto-segmentation for each organ at risk (OAR) in the thorax.MethodsA radiation oncologist manually delineated 13 thoracic OARs from computed tomography (CT) images of 40 patients. OAR auto-segmentation accuracy was graded by five geometric objective indexes, including the Dice similarity coefficient (DSC), the difference of the Euclidean distance between centers of mass (ΔCMD), the difference of volume (ΔV), maximum Hausdorff distance (MHD), and average Hausdorff distance (AHD). The grading results were compared with those of the corresponding geometric indexes obtained by geometric objective methods in the other two centers. OAR auto-segmentation accuracy was also graded by our subjective evaluation standard. These grading results were compared with those of DSC. Based on the subjective evaluation standard and the five geometric indexes, the correspondence between the subjective evaluation level and the geometric index range was established for each OAR.ResultsFor ΔCMD, ΔV, and MHD, the grading results of the geometric objective evaluation methods at our center and the other two centers were inconsistent. For DSC and AHD, the grading results of three centers were consistent. Seven OARs’ grading results in the subjective evaluation standard were inconsistent with those of DSC. Six OARs’ grading results in the subjective evaluation standard were consistent with those of DSC. Finally, we proposed a new evaluation method that combined the subjective evaluation level of those OARs with the range of corresponding DSC to determine the grading standard. If the DSC ranges between the adjacent levels did not overlap, the DSC range was used as the grading standard. Otherwise, the mean value of DSC was used as the grading standard.ConclusionsA novel OAR-specific SOC grading standard in thorax was developed. The SOC grading standard provides a possible alternative for evaluation of the auto-segmentation accuracy for thoracic OARs.

Automated measurement of sulcus angle on axial knee magnetic resonance images

AbstractPatellar dislocation is a disorder in the human knee where patella slips out of its expected position. The orthopedic experts need to manually measure the parameters indicating patellar instability from knee scans, which is a laborious and time‐consuming task. An automated method for measuring these parameters can resolve the issue of inter‐observer and intra‐observer variations and significantly lessen the burden of the experts. Therefore, in this work, a two‐step approach has been proposed for the automated measurement of the sulcus angle from knee magnetic resonance (MR) images to assist the experts in the diagnosis of patellar dislocation. Firstly, a variant of U‐Net architecture has been proposed harnessing the capabilities of the residual network to segment the region of interest (ROI) from the axial knee MR images. Secondly, the contour of ROI is used for the detection of key points required to compute sulcus angle. A T2‐weighted knee magnetic resonance imaging dataset of 50 patients obtained from Post Graduate Institute of Medical Education &amp; Research, Chandigarh, has been used. Out of 50 patients, the randomly selected knee MR images of 25 patients have been used to train the segmentation model. The intersection over union and dice similarity coefficient values of 94.71% and 97.28%, respectively, were achieved for the proposed segmentation model. The automated measurement of the sulcus angle resulted in a mean error of 2.52 ± 1.52° when tested on another set of axial knee MR images of 25 patients. The proposed framework produced results comparable to the manual expert measurements.

3D brain glioma segmentation in MRI through integrating multiple densely connected 2D convolutional neural networks.

To overcome the computational burden of processing three-dimensional (3D) medical scans and the lack of spatial information in two-dimensional (2D) medical scans, a novel segmentation method was proposed that integrates the segmentation results of three densely connected 2D convolutional neural networks (2D-CNNs). In order to combine the low-level features and high-level features, we added densely connected blocks in the network structure design so that the low-level features will not be missed as the network layer increases during the learning process. Further, in order to resolve the problems of the blurred boundary of the glioma edema area, we superimposed and fused the T2-weighted fluid-attenuated inversion recovery (FLAIR) modal image and the T2-weighted (T2) modal image to enhance the edema section. For the loss function of network training, we improved the cross-entropy loss function to effectively avoid network over-fitting. On the Multimodal Brain Tumor Image Segmentation Challenge (BraTS) datasets, our method achieves dice similarity coefficient values of 0.84, 0.82, and 0.83 on the BraTS2018 training; 0.82, 0.85, and 0.83 on the BraTS2018 validation; and 0.81, 0.78, and 0.83 on the BraTS2013 testing in terms of whole tumors, tumor cores, and enhancing cores, respectively. Experimental results showed that the proposed method achieved promising accuracy and fast processing, demonstrating good potential for clinical medicine.