Mean Dice Similarity Coefficient Research Articles

A deep learning-based method for the detection and segmentation of breast masses in ultrasound images.

Automated detection and segmentation of breast masses in ultrasound images are critical for breast cancer diagnosis, but remain challenging due to limited image quality and complex breast tissues. This study aims to develop a deep learning-based method that enables accurate breast mass detection and segmentation in ultrasound images.&#xD;Approach. A novel convolutional neural network-based framework that combines the You Only Look Once (YOLO) v5 network and the Global-Local (GOLO) strategy was developed. First, YOLOv5 was applied to locate the mass regions of interest (ROIs). Second, a Global Local-Connected Multi-Scale Selection (GOLO-CMSS) network was developed to segment the masses. The GOLO-CMSS operated on both the entire images globally and mass ROIs locally, and then integrated the two branches for a final segmentation output. Particularly, in global branch, CMSS applied Multi-Scale Selection (MSS) modules to automatically adjust the receptive fields, and Multi-Input (MLI) modules to enable fusion of shallow and deep features at different resolutions. The USTC dataset containing 28,477 breast ultrasound images was collected for training and test. The proposed method was also tested on three public datasets, UDIAT, BUSI and TUH. The segmentation performance of GOLO-CMSS was compared with others networks and three experienced radiologists.&#xD;Main results. YOLOv5 outperformed other detection models with average precisions of 99.41%, 95.15%, 93.69% and 96.42% on the USTC, UDIAT, BUSI and TUH datasets, respectively. The proposed GOLO-CMSS showed superior segmentation performance over other state-of-the-art networks, with Dice similarity coefficients (DSCs) of 93.19%, 88.56%, 87.58% and 90.37% on the USTC, UDIAT, BUSI and TUH datasets, respectively. The mean DSC between GOLO-CMSS and each radiologist was significantly better than that between radiologists (p < 0.001).&#xD;Significance. Our proposed method can accurately detect and segment breast masses with a decent performance comparable to radiologists, highlighting its great potential for clinical implementation in breast ultrasound examination.

ML-driven segmentation of microvascular features during histological examination of tissue-engineered vascular grafts.

The development of next-generation tissue-engineered medical devices such as tissue-engineered vascular grafts (TEVGs) is a leading trend in translational medicine. Microscopic examination is an indispensable part of animal experimentation, and histopathological analysis of regenerated tissue is crucial for assessing the outcomes of implanted medical devices. However, the objective quantification of regenerated tissues can be challenging due to their unusual and complex architecture. To address these challenges, research and development of advanced ML-driven tools for performing adequate histological analysis appears to be an extremely promising direction. We compiled a dataset of 104 representative whole slide images (WSIs) of TEVGs which were collected after a 6-month implantation into the sheep carotid artery. The histological examination aimed to analyze the patterns of vascular tissue regeneration in TEVGs in situ. Having performed an automated slicing of these WSIs by the Entropy Masker algorithm, we filtered and then manually annotated 1,401 patches to identify 9 histological features: arteriole lumen, arteriole media, arteriole adventitia, venule lumen, venule wall, capillary lumen, capillary wall, immune cells, and nerve trunks. To segment and quantify these features, we rigorously tuned and evaluated the performance of six deep learning models (U-Net, LinkNet, FPN, PSPNet, DeepLabV3, and MA-Net). After rigorous hyperparameter optimization, all six deep learning models achieved mean Dice Similarity Coefficients (DSC) exceeding 0.823. Notably, FPN and PSPNet exhibited the fastest convergence rates. MA-Net stood out with the highest mean DSC of 0.875, demonstrating superior performance in arteriole segmentation. DeepLabV3 performed well in segmenting venous and capillary structures, while FPN exhibited proficiency in identifying immune cells and nerve trunks. An ensemble of these three models attained an average DSC of 0.889, surpassing their individual performances. This study showcases the potential of ML-driven segmentation in the analysis of histological images of tissue-engineered vascular grafts. Through the creation of a unique dataset and the optimization of deep neural network hyperparameters, we developed and validated an ensemble model, establishing an effective tool for detecting key histological features essential for understanding vascular tissue regeneration. These advances herald a significant improvement in ML-assisted workflows for tissue engineering research and development.

Impact of gadopiclenol on decision making in patients with brain metastases: A post-hoc study.

e14003 Background: Gadopiclenol (Elucirem, Guerbet, France) is a new high relaxivity macrocyclic gadolinium-based contrast agent (GBCA), recently approved by the FDA and EMA at the dose of 0.05 mmol/kg. The aim of this post-hoc analysis was to evaluate the impact of contrast-enhanced MRI with gadopiclenol at 0.1 mmol/kg on decision making and radiotherapy (RT) treatment planning of brain metastases (BM). Methods: This is a post-hoc analysis of data from a phase IIb study where patients underwent two separate MRI examinations, one with gadopiclenol at 0.025, 0.05, 0.1 or 0.2 mmol/kg and one with gadobenate dimeglumine at 0.1 mmol/kg. MR images of patients who received both GBCAs at 0.1 mmol/kg, with ≥1 BM detected in either of the scans, were subjected to a blinded reader analysis and contouring by two radiation oncology experts. For each patient, treatment plans (stereotactic radiosurgery [SRS] or whole-brain radiotherapy [WBRT]) were determined for both MRIs, with the gross target volume (GTV) indicating the contrast-enhancing aspects of the tumor. Mean GTVs and normal tissue volumes receiving 12 Gy (V12), as well as the Dice similarity coefficient (DSC) were obtained for the paired contours. The Spearman´s rank (ρ) correlation was additionally calculated. Furthermore, three experts blindly evaluated the contrast enhancement of each lesion for contouring purposes and subjectively qualified them as “better”, in detriment of the counterpart, or “equal”. Results: In total, images from 13 adult patients were analyzed. MRI with gadopiclenol depicted additional BM as compared with gadobenate dimeglumine in 7 patients (54%). The treatment plan was changed in 2 patients (15%), from no treatment to SRS and from SRS to WBRT. Gadopiclenol depicted additional BM in these 2 patients (from 0 and 10 BM with gadobenate dimeglumine to 1 and 15 BM with gadopiclenol). The mean GTVs and V12 were comparable between gadopiclenol and gadobenate dimeglumine (p=0.694, p=1.974). The mean DSC was 0.70 (SD: 0.14, ρ0.82). From a total of 36 answers, contrast enhancement was qualified as better with gadopiclenol in 21 (58.3%) evaluations, better with gadobenate dimeglumine in 8 (22.2%) evaluations, while no difference was observed in 7 (19.4%) evaluations. Conclusions: Gadopiclenol at 0.1 mmol/kg improved BM detection and contrast enhancement with potential impact on RT treatment decisions.

Diffeomorphic transformer-based abdomen MRI-CT deformable image registration.

Stereotactic body radiotherapy (SBRT) is a well-established treatment modality for liver metastases in patients unsuitable for surgery. Both CT and MRI are useful during treatment planning for accurate target delineation and to reduce potential organs-at-risk (OAR) toxicity from radiation. MRI-CT deformable image registration (DIR) is required to propagate the contours defined on high-contrast MRI to CT images. An accurate DIR method could lead to more precisely defined treatment volumes and superior OAR sparing on the treatment plan. Therefore, it is beneficial to develop an accurate MRI-CT DIR for liver SBRT. To create a new deep learning model that can estimate the deformation vector field (DVF) for directly registering abdominal MRI-CT images. The proposed method assumed a diffeomorphic deformation. By using topology-preserved deformation features extracted from the probabilistic diffeomorphic registration model, abdominal motion can be accurately obtained and utilized for DVF estimation. The model integrated Swin transformers, which have demonstrated superior performance in motion tracking, into the convolutional neural network (CNN) for deformation feature extraction. The model was optimized using a cross-modality image similarity loss and a surface matching loss. To compute the image loss, a modality-independent neighborhood descriptor (MIND) was used between the deformed MRI and CT images. The surface matching loss was determined by measuring the distance between the warped coordinates of the surfaces of contoured structures on the MRI and CT images. To evaluate the performance of the model, a retrospective study was carried out on a group of 50 liver cases that underwent rigid registration of MRI and CT scans. The deformed MRI image was assessed against the CT image using the target registration error (TRE), Dice similarity coefficient (DSC), and mean surface distance (MSD) between the deformed contours of the MRI image and manual contours of the CT image. When compared to only rigid registration, DIR with the proposed method resulted in an increase of the mean DSC values of the liver and portal vein from 0.850±0.102 and 0.628±0.129 to 0.903±0.044 and 0.763±0.073, a decrease of the mean MSD of the liver from 7.216±4.513mm to 3.232±1.483mm, and a decrease of the TRE from 26.238±2.769mm to 8.492±1.058mm. The proposed DIR method based on a diffeomorphic transformer provides an effective and efficient way to generate an accurate DVF from an MRI-CT image pair of the abdomen. It could be utilized in the current treatment planning workflow for liver SBRT.

Neural network dose prediction for cervical brachytherapy: Overcoming data scarcity for applicator-specific models.

3D neural network dose predictions are useful for automating brachytherapy (BT) treatment planning for cervical cancer. Cervical BT can be delivered with numerous applicators, which necessitates developing models that generalize to multiple applicator types. The variability and scarcity of data for any given applicator type poses challenges for deep learning. The goal of this work was to compare three methods of neural network training-a single model trained on all applicator data, fine-tuning the combined model to each applicator, and individual (IDV) applicator models-to determine the optimal method for dose prediction. Models were produced for four applicator types-tandem-and-ovoid (T&O), T&O with 1-7 needles (T&ON), tandem-and-ring (T&R) and T&R with 1-4 needles (T&RN). First, the combined model was trained on 859 treatment plans from 266 cervical cancer patients treated from 2010 onwards. The train/validation/test split was 70%/16%/14%, with approximately 49%/10%/19%/22% T&O/T&ON/T&R/T&RN in each dataset. Inputs included four channels for anatomical masks (high-risk clinical target volume [HRCTV], bladder, rectum, and sigmoid), a mask indicating dwell position locations, and applicator channels for each applicator component. Applicator channels were created by mapping the 3D dose for a single dwell position to each dwell position and summing over each applicator component with uniform dwell time weighting. A 3D Cascade U-Net, which consists of two U-Nets in sequence, and mean squared error loss function were used. The combined model was then fine-tuned to produce four applicator-specific models by freezing the first U-Net and encoding layers of the second and resuming training on applicator-specific data. Finally, four IDV models were trained using only data from each applicator type. Performance of these three model types was compared using the following metrics for the test set: mean error (ME, representing model bias) and mean absolute error (MAE) over all dose voxels and ME of clinical metrics (HRCTV D90% and D2cc of bladder, rectum, and sigmoid), averaged over all patients. A positive ME indicates the clinical dose was higher than predicted. 3D global gamma analysis with the prescription dose as reference value was performed. Dice similarity coefficients (DSC) were computed for each isodose volume. Fine-tuned and combined models showed better performance than IDV applicator training. Fine-tuning resulted in modest improvements in about half the metrics, compared to the combined model, while the remainder were mostly unchanged. Fine-tuned MAE=3.98%/2.69%/5.36%/3.80% for T&O/T&R/T&ON/T&RN, and ME over all voxels=-0.08%/-0.89%/-0.59%/1.42%. ME D2cc were bladder=-0.77%/1.00%/-0.66%/-1.53%, rectum=1.11%/-0.22%/-0.29%/-3.37%, sigmoid=-0.47%/-0.06%/-2.37%/-1.40%, and ME D90=2.6%/-4.4%/4.8%/0.0%. Gamma pass rates (3%/3mm) were 86%/91%/83%/89%. Mean DSCs were 0.92%/0.92%/0.88%/0.91% for isodoses ≤150% of prescription. 3D BT dose was accurately predicted for all applicator types, as indicated by the low MAE and MEs, high gamma scores and high DSCs. Training on all treatment data overcomes challenges with data scarcity in each applicator type, resulting in superior performance than can be achieved by training on IDV applicators alone. This could presumably be explained by the fact that the larger, more diverse dataset allows the neural network to learn underlying trends and characteristics in dose that are common to all treatment applicators. Accurate, applicator-specific dose predictions could enable automated, knowledge-based planning for any cervical brachytherapy treatment.

Automated deep learning segmentation of neuritic plaques and neurofibrillary tangles in Alzheimer disease brain sections using a proprietary software.

Neuropathological diagnosis of Alzheimer disease (AD) relies on semiquantitative analysis of phosphorylated tau-positive neurofibrillary tangles (NFTs) and neuritic plaques (NPs), without consideration of lesion heterogeneity in individual cases. We developed a deep learning workflow for automated annotation and segmentation of NPs and NFTs from AT8-immunostained whole slide images (WSIs) of AD brain sections. Fifteen WSIs of frontal cortex from 4 biobanks with varying tissue quality, staining intensity, and scanning formats were analyzed. We established an artificial intelligence (AI)-driven iterative procedure to improve the generation of expert-validated annotation datasets for NPs and NFTs thereby increasing annotation quality by >50%. This strategy yielded an expert-validated annotation database with 5013 NPs and 5143 NFTs. We next trained two U-Net convolutional neural networks for detection and segmentation of NPs or NFTs, achieving high accuracy and consistency (mean Dice similarity coefficient: NPs, 0.77; NFTs, 0.81). The workflow showed high generalization performance across different cases. This study serves as a proof-of-concept for the utilization of proprietary image analysis software (Visiopharm) in the automated deep learning segmentation of NPs and NFTs, demonstrating that AI can significantly improve the annotation quality of complex neuropathological features and enable the creation of highly precise models for identifying these markers in AD brain sections.

HCA-DAN: hierarchical class-aware domain adaptive network for gastric tumor segmentation in 3D CT images

BackgroundAccurate segmentation of gastric tumors from CT scans provides useful image information for guiding the diagnosis and treatment of gastric cancer. However, automated gastric tumor segmentation from 3D CT images faces several challenges. The large variation of anisotropic spatial resolution limits the ability of 3D convolutional neural networks (CNNs) to learn features from different views. The background texture of gastric tumor is complex, and its size, shape and intensity distribution are highly variable, which makes it more difficult for deep learning methods to capture the boundary. In particular, while multi-center datasets increase sample size and representation ability, they suffer from inter-center heterogeneity.MethodsIn this study, we propose a new cross-center 3D tumor segmentation method named Hierarchical Class-Aware Domain Adaptive Network (HCA-DAN), which includes a new 3D neural network that efficiently bridges an Anisotropic neural network and a Transformer (AsTr) for extracting multi-scale context features from the CT images with anisotropic resolution, and a hierarchical class-aware domain alignment (HCADA) module for adaptively aligning multi-scale context features across two domains by integrating a class attention map with class-specific information. We evaluate the proposed method on an in-house CT image dataset collected from four medical centers and validate its segmentation performance in both in-center and cross-center test scenarios.ResultsOur baseline segmentation network (i.e., AsTr) achieves best results compared to other 3D segmentation models, with a mean dice similarity coefficient (DSC) of 59.26%, 55.97%, 48.83% and 67.28% in four in-center test tasks, and with a DSC of 56.42%, 55.94%, 46.54% and 60.62% in four cross-center test tasks. In addition, the proposed cross-center segmentation network (i.e., HCA-DAN) obtains excellent results compared to other unsupervised domain adaptation methods, with a DSC of 58.36%, 56.72%, 49.25%, and 62.20% in four cross-center test tasks.ConclusionsComprehensive experimental results demonstrate that the proposed method outperforms compared methods on this multi-center database and is promising for routine clinical workflows.

Deep-Learning Models for Abdominal CT Organ Segmentation in Children: Development and Validation in Internal and Heterogeneous Public Datasets.

Background: Deep-learning abdominal organ segmentation algorithms have shown excellent results in adults; validation in children is sparse. Objective: To develop and validate deep-learning models for liver, spleen, and pancreas segmentation on pediatric CT examinations. Methods: This retrospective study developed and validated deep-learning models for liver, spleen, and pancreas segmentation using 1731 CT examinations (1504 training, 221 testing), derived from three internal institutional pediatric (age ≤18) datasets (n=483) and three public datasets comprising pediatric and adult examinations with various pathologies (n=1248). Three deep-learning model architectures (SegResNet, DynUNet, and SwinUNETR) from the Medical Open Network for AI (MONAI) framework underwent training using native training (NT), relying solely on institutional datasets, and transfer learning (TL), incorporating pre-training on public datasets. For comparison, TotalSegmentator (TS), a publicly available segmentation model, was applied to test data without further training. Segmentation performance was evaluated using mean Dice similarity coefficient (DSC), with manual segmentations as reference. Results: For internal pediatric data, DSC for normal liver was 0.953 (TS), 0.964-0.965 (NT models), and 0.965-0.966 (TL models); normal spleen, 0.914 (TS), 0.942-0.945 (NT models), and 0.937-0.945 (TL models); normal pancreas, 0.733 (TS), 0.774-0.785 (NT models), and 0.775-0.786 (TL models); pancreas with pancreatitis, 0.703 (TS), 0.590-0.640 (NT models), and 0.667-0.711 (TL models). For public pediatric data, DSC for liver was 0.952 (TS), 0.876-0.908 (NT models), and 0.941-0.946 (TL models); spleen, 0.905 (TS), 0.771-0.827 (NT models), and 0.897-0.926 (TL models); pancreas, 0.700 (TS), 0.577-0.648 (NT models), and 0.693-0.736 (TL models). For public primarily adult data, DSC for liver was 0.991 (TS), 0.633-0.750 (NT models), and 0.926-0.952 (TL models); spleen, 0.983 (TS), 0.569-0.604 (NT models), and 0.923-0.947 (TL models); pancreas, 0.909 (TS), 0.148-0.241 (NT models), and 0.699-0.775 (TL models). DynUNet-TL was selected as the best-performing NT or TL model and was made available as an opensource MONAI bundle (https://github.com/cchmc-dll/pediatric_abdominal_segmentation_bundle.git). Conclusion: TL models trained on heterogeneous public datasets and fine-tuned using institutional pediatric data outperformed internal NT models and TotalSegmentator across internal and external pediatric test data. Segmentation performance was better in liver and spleen than in pancreas. Clinical Impact: The selected model may be used for various volumetry applications in pediatric imaging.

Semi-supervised learning in diagnosis of infant hip dysplasia towards multisource ultrasound images.

Automated diagnosis of infant hip dysplasia is heavily affected by the individual differences among infants and ultrasound machines. Hip sonographic images of 493 infants from various ultrasound machines were collected in the Department of Orthopedics in Yangzhou Maternal and Child Health Care Service Centre. Herein, we propose a semi-supervised learning method based on a feature pyramid network (FPN) and a contrastive learning scheme based on a Siamese architecture. A large amount of unlabeled data of ultrasound images was used via the Siamese network in the pre-training step, and then a small amount of annotated data for anatomical structures was adopted to train the model for landmark identification and standard plane recognition. The method was evaluated on our collected dataset. The method achieved a mean Dice similarity coefficient (DSC) of 0.7873 and a mean Hausdorff distance (HD) of 5.0102 in landmark identification, compared to the model without contrastive learning, which had a mean DSC of 0.7734 and a mean HD of 6.1586. The accuracy, precision, and recall of standard plane recognition were 95.4%, 91.64%, and 94.86%, respectively. The corresponding area under the curve (AUC) was 0.982. This study proposes a semi-supervised deep learning method following Graf's principle, which can better utilize a large volume of ultrasound images from various devices and infants. This method can identify the landmarks of infant hips more accurately than manual operators, thereby improving the efficiency of diagnosis of infant hip dysplasia.

A deep learning-based 3D Prompt-nnUnet model for automatic segmentation in brachytherapy of postoperative endometrial carcinoma.

To create and evaluate a three-dimensional (3D) Prompt-nnUnet module that utilizes the prompts-based model combined with 3D nnUnet for producing the rapid and consistent autosegmentation of high-risk clinical target volume (HR CTV) and organ at risk (OAR) in high-dose-rate brachytherapy (HDR BT) for patients with postoperative endometrial carcinoma (EC). On two experimental batches, a total of 321 computed tomography (CT) scans were obtained for HR CTV segmentation from 321 patients with EC, and 125 CT scans for OARs segmentation from 125 patients. The numbers of training/validation/test were 257/32/32 and 87/13/25 for HR CTV and OARs respectively. A novel comparison of the deep learning neural network 3D Prompt-nnUnet and 3D nnUnet was applied for HR CTV and OARs segmentation. Three-fold cross validation and several quantitative metrics were employed, including Dice similarity coefficient (DSC), Hausdorff distance (HD), 95th percentile of Hausdorff distance (HD95%), and intersection over union (IoU). The Prompt-nnUnet included two forms of parameters Predict-Prompt (PP) and Label-Prompt (LP), with the LP performing most similarly to the experienced radiation oncologist and outperforming the less experienced ones. During the testing phase, the mean DSC values for the LP were 0.96±0.02, 0.91±0.02, and 0.83±0.07 for HR CTV, rectum and urethra, respectively. The mean HD values (mm) were 2.73±0.95, 8.18±4.84, and 2.11±0.50, respectively. The mean HD95% values (mm) were 1.66±1.11, 3.07±0.94, and 1.35±0.55, respectively. The mean IoUs were 0.92±0.04, 0.84±0.03, and 0.71±0.09, respectively. A delineation time<2.35s per structure in the new model was observed, which was available to save clinician time. The Prompt-nnUnet architecture, particularly the LP, was highly consistent with ground truth (GT) in HR CTV or OAR autosegmentation, reducing interobserver variability and shortening treatment time.

Comparison of commercial atlas-based automatic segmentation software for prostate radiotherapy treatment planning.

This study aims to assess the accuracy of automatic atlas-based contours for various key anatomical structures in prostate radiotherapy treatment planning. The evaluated structures include the bladder, rectum, prostate, seminal vesicles, femoral heads and penile bulb. CT images from 20 patients who underwent intensity-modulated radiotherapy were randomly chosen to create an atlas library. Atlas contours of the seven anatomical structures were generated using four software packages: ABAS, Eclipse, MIM, and RayStation. These contours were then compared to manual delineations performed by oncologists, which served as the ground truth. Evaluation metrics such as dice similarity coefficient (DSC), mean distance to agreement (MDA), and volume ratio (VR) were calculated to assess the accuracy of the contours. Additionally, the time taken by each software to generate the atlas contour was recorded. The mean DSC values for the bladder exhibited strong agreement (>0.8) with manual delineations for all software except for Eclipse and RayStation. Similarly, the femoral heads showed significant similarity between the atlas contours and ground truth across all software, with mean DSC values exceeding 0.9 and MDA values close to zero. On the other hand, the penile bulb displayed only moderate agreement with the ground truth, with mean DSC values ranging from 0.5 to 0.7 for all software. A similar trend was observed in the prostate atlas contours, except for MIM, which achieved a mean DSC of over 0.8. For the rectum, both ABAS and MIM atlases demonstrated strong agreement with the ground truth, resulting in mean DSC values of more than 0.8. Overall, MIM and ABAS outperformed Eclipse and RayStation in both DSC and MDA. These results indicate that the atlas-based segmentation employed in this study produces acceptable contours for the anatomical structures of interest in prostate radiotherapy treatment planning.

Improving hybrid image and structure-based deformable image registration for large internal deformations

Objective. Deformable image registration (DIR) is a widely used technique in radiotherapy. Complex deformations, resulting from large anatomical changes, are a regular challenge. DIR algorithms generally seek a balance between capturing large deformations and preserving a smooth deformation vector field (DVF). We propose a novel structure-based term that can enhance the registration efficacy while ensuring a smooth DVF. Approach. The proposed novel similarity metric for controlling structures was introduced as a new term into a commercially available algorithm. Its performance was compared to the original algorithm using a dataset of 46 patients who received pelvic re-irradiation, many of which exhibited complex deformations. Main results. The mean Dice Similarity Coefficient (DSC) under the improved algorithm was 0.96, 0.94, 0.76, and 0.91 for bladder, rectum, colon, and bone respectively, compared to 0.69, 0.89, 0.62, and 0.88 for the original algorithm. The improvement was more pronounced for complex deformations. Significance. With this work, we have demonstrated that the proposed term is able to improve registration accuracy for complex cases while maintaining realistic deformations.

Is 3T MR nerve-bone fusion imaging a viable alternative to MRI-CBCT to identify the relationship between the inferior alveolar nerve and mandibular third molar.

To investigate the feasibility of MRI nerve-bone fusion imaging in assessing the relationship between inferior alveolar nerve (IAN) / mandibular canal (MC) and mandibular third molar (MTM) compared with MRI-CBCT fusion. The MRI nerve-bone fusion and MRI-CBCT fusion imaging were performed in 20 subjects with 37 MTMs. The Hausdorff distance (HD) value and dice similarity coefficient (DSC) was calculated. The relationship between IAN/MC and MTM roots, inflammatory, and fusion patterns were compared between these two fused images. The reliability was assessed using a weighted κ statistic. The mean HD and DSC ranged from 0.62~1.35 and 0.83~0.88 for MRI nerve-bone fusion, 0.98 ~ 1.50 and 0.76 ~ 0.83 for MRI-CBCT fusion. MR nerve-bone fusion had considerable reproducibility compared to MRI-CBCT fusion in relation classification (MR nerve-bone fusion κ = 0.694, MRI-CBCT fusion κ = 0.644), direct contact (MR nerve-bone fusion κ = 0.729, MRI-CBCT fusion κ = 0.720), and moderate to good agreement for inflammation detection (MR nerve-bone fusion κ = 0.603, MRI-CBCT fusion κ = 0.532, average). The MR nerve-bone fusion imaging showed a lower ratio of larger pattern compared to MR-CBCT fusion (16.2% VS 27.3% in the molar region, and 2.7% VS 5.4% in the retromolar region). And the average time spent on MR nerve-bone fusion and MRI-CBCT fusion was 1min and 3min, respectively. Both MR nerve-bone fusion and MRI-CBCT fusion exhibited good consistency in evaluating the spatial relationship between IAN/MC and MTM, fusion effect, and inflammation detection. MR nerve-bone fusion imaging can be a preoperative one-stop radiation-free examination for patients at high risk for MTM surgery.

Automatic segmentation of hemorrhagic transformation on follow-up non-contrast CT after acute ischemic stroke.

Hemorrhagic transformation (HT) following reperfusion therapies is a serious complication for patients with acute ischemic stroke. Segmentation and quantification of hemorrhage provides critical insights into patients' condition and aids in prognosis. This study aims to automatically segment hemorrhagic regions on follow-up non-contrast head CT (NCCT) for stroke patients treated with endovascular thrombectomy (EVT). Patient data were collected from 10 stroke centers across two countries. We propose a semi-automated approach with adaptive thresholding methods, eliminating the need for extensive training data and reducing computational demands. We used Dice Similarity Coefficient (DSC) and Lin's Concordance Correlation Coefficient (Lin's CCC) to evaluate the performance of the algorithm. A total of 51 patients were included, with 28 Type 2 hemorrhagic infarction (HI2) cases and 23 parenchymal hematoma (PH) cases. The algorithm achieved a mean DSC of 0.66 ± 0.17. Notably, performance was superior for PH cases (mean DSC of 0.73 ± 0.14) compared to HI2 cases (mean DSC of 0.61 ± 0.18). Lin's CCC was 0.88 (95% CI 0.79-0.93), indicating a strong agreement between the algorithm's results and the ground truth. In addition, the algorithm demonstrated excellent processing time, with an average of 2.7 s for each patient case. To our knowledge, this is the first study to perform automated segmentation of post-treatment hemorrhage for acute stroke patients and evaluate the performance based on the radiological severity of HT. This rapid and effective tool has the potential to assist with predicting prognosis in stroke patients with HT after EVT.

3D auto-segmentation of biliary structure of living liver donors using magnetic resonance cholangiopancreatography for enhanced preoperative planning.

This study aimed to develop an automated segmentation system for biliary structures using a deep learning model, based on data from magnetic resonance cholangiopancreatography (MRCP). Living liver donors who underwent MRCP using the gradient and spin echo technique followed by three-dimensional modeling were eligible for this study. A three-dimensional residual U-Net model was implemented for the deep learning process. Data were divided into training and test sets at a 9:1 ratio. Performance was assessed using the dice similarity coefficient to compare the model's segmentation with the manually labeled ground truth. The study incorporated 250 cases. There was no difference in the baseline characteristics between the train set (n=225) and test set (n=25). The overall mean Dice Similarity Coefficient was 0.80±0.20 between the ground truth and inference result. The qualitative assessment of the model showed relatively high accuracy especially for the common bile duct (88%), common hepatic duct (92%), hilum (96%), right hepatic duct (100%), and left hepatic duct (96%), while the third-order branch of the right hepatic duct (18.2%) showed low accuracy. The developed automated segmentation model for biliary structures, utilizing MRCP data and deep learning techniques, demonstrated robust performance and holds potential for further advancements in automation.

A deep learning model for translating CT to ventilation imaging: analysis of accuracy and impact on functional avoidance radiotherapy planning.

Radiotherapy planning incorporating functional lung images has the potential to reduce pulmonary toxicity. Free-breathing 4DCT-derived ventilation image (CTVI) may help quantify lung function. This study introduces a novel deep-learning model directly translating planning CT images into CTVI. We investigated the accuracy of generated images and the impact on functional avoidance planning. Paired planning CT and 4DCT scans from 48 patients with NSCLC were randomized to training (n = 41) and testing (n = 7) data sets. The ventilation maps were generated from 4DCT using a Jacobian-based algorithm to provide ground truth labels (CTVI4DCT). A 3D U-Net-based model was trained to map CT to synthetic CTVI (CTVISyn) and validated using fivefold cross-validation. The highest-performing model was applied to the testing set. Spearman's correlation (rs) and Dice similarity coefficient (DSC) determined voxel-wise and functional-wise concordance between CTVI4DCT and CTVISyn. Three plans were designed per patient in the testing set: one clinical plan without CTVI and two functional avoidance plans combined with CTVI4DCT or CTVISyn, aimed at sparing high-functional lungs defined as the top 50% of the percentile ventilation ranges. Dose-volume (DVH) parameters regarding the planning target volume (PTV) and organs at risk (OARs) were recorded. Radiation pneumonitis (RP) risk was estimated using a dose-function (DFH)-based normal tissue complication probability (NTCP) model. CTVISyn showed a mean rs value of 0.65 ± 0.04 compared to CTVI4DCT. Mean DSC values over the top 50% and 60% of ventilation ranges were 0.41 ± 0.07 and 0.52 ± 0.10, respectively. In the test set (n = 7), all patients' RP-risk benefited from CTVI4DCT-guided plans (Riskmean_4DCT_vs_Clinical: 29.24% vs. 49.12%, P = 0.016), and six patients benefited from CTVISyn-guided plans (Riskmean_Syn_vs_Clinical: 31.13% vs. 49.12%, P = 0.022). There were no significant differences in DVH and DFH metrics between CTVISyn and CTVI4DCT-guided plan (P > 0.05). Using deep-learning techniques, CTVISyn generated from planning CT exhibited a moderate-to-high correlation with CTVI4DCT. The CTVISyn-guided plans were comparable to the CTVI4DCT-guided plans, effectively reducing pulmonary toxicity in patients while maintaining acceptable plan quality. Further prospective trials are needed to validate these findings.

Quantitative measurement of the ureter on three-dimensional magnetic resonance urography images using deep learning.

Accurate measurement of ureteral diameters plays a pivotal role in diagnosing and monitoring urinary tract obstruction (UTO). While three-dimensional magnetic resonance urography (3D MRU) represents a significant advancement in imaging, the traditional manual methods for assessing ureteral diameters are characterized by labor-intensive procedures and inherent variability. In the realm of medical image analysis, deep learning has led to a paradigm shift, yet the development of a comprehensive automated tool for the precise segmentation and measurement of ureters in MR images is an unaddressed challenge. The ureter was quantitatively measured on 3D MRU images using a deep learning model. A retrospective cohort of 445 3D MRU scans (443 patients, 52±18 years; 217 female patients) was collected and split into training, validation, and internal testing cohorts. A 3D V-Net model was trained for urinary tract segmentation, and a post-processing algorithm was developed for ureteral measurements. The accuracy of the segmentation was evaluated using the Dice similarity coefficient (DSC) and volume intraclass correlation coefficient (ICC), with ground truth segmentations provided by experienced radiologists. The external cohort comprised 50 scans (50 patients, 55±21 years; 30 female patients), and the model-predicted ureteral diameter measurements were compared with manual measurements to assess system performance. The various diameter parameters of ureter among the different measurement methods (ground truth, auto-segmentation with automatic diameter extraction, and manual segmentation with automatic diameter extraction) were assessed with Friedman tests and post hoc Dunn test. The effectiveness of the UTO diagnosis was assessed by receiver operating characteristic (ROC) curves and their respective areas under the curve (AUC) between different methods. In both the internal test and external cohorts, the mean DSC values for bilateral ureters exceeded 0.70. The ICCs for the bilateral ureter volume obtained by comparing the model and manual segmentation were all greater than 0.96 (p < 0.05), except for the right ureter in the internal test cohort, for which the ICC was 0.773 (p < 0.05). The mean DSCs for interobserver and intraobserver reliability were all above 0.97. The maximum diameter of the ureter exhibited no statistically significant differences either in the dilated (p=0.08) or in the non-dilated (p=0.32) ureters across the three measurement methods. The AUCs of ground truth, auto-segmentation with automatic diameter extraction, and manual segmentation with automatic diameter extraction in diagnosing UTO were 0.988 (95% CI: 0.934, 1.000), 0.961 (95% CI: 0.893, 0.991), and 0.979 (95% CI: 0.919, 0.998), respectively. There was no statistical difference between AUCs of the different methods (p>0.05). The proposed deep learning model and post-processing algorithm provide an effective means for the quantitative evaluation of urinary diseases using 3D MRU images.

Dual-branch feature fusion S3D V-Net network for lung nodules segmentation.

Accurate segmentation of lung nodules can help doctors get more accurate results and protocols in early lung cancer diagnosis and treatment planning, so that patients can be better detected and treated at an early stage, and the mortality rate of lung cancer can be reduced. Currently, the improvement of lung nodule segmentation accuracy has been limited by his heterogeneous performance in the lungs, the imbalance between segmentation targets and background pixels, and other factors. We propose a new 2.5D lung nodule segmentation network model for lung nodule segmentation. This network model can well improve the extraction of edge information of lung nodules, and fuses intra-slice and inter-slice features, which makes good use of the three-dimensional structural information of lung nodules and can more effectively improve the accuracy of lung nodule segmentation. Our approach is based on a typical encoding-decoding network structure for improvement. The improved model captures the features of multiple nodules in both 3-D and 2-D CT images, complements the information of the segmentation target's features and enhances the texture features at the edges of the pulmonary nodules through the dual-branch feature fusion module (DFFM) and the reverse attention context module (RACM), and employs central pooling instead of the maximal pooling operation, which is used to preserve the features around the target and to eliminate the edge-irrelevant features, to further improve the performance of the segmentation of the pulmonary nodules. We evaluated this method on a wide range of 1186 nodules from the LUNA16 dataset, and averaging the results of ten cross-validated, the proposed method achieved the mean dice similarity coefficient (mDSC) of 84.57%, the mean overlapping error (mOE) of 18.73% and average processing of a case is about 2.07 s. Moreover, our results were compared with inter-radiologist agreement on the LUNA16 dataset, and the average difference was 0.74%. The experimental results show that our method improves the accuracy of pulmonary nodules segmentation and also takes less time than more 3-D segmentation methods in terms of time.

Automatic segmentation of hepatocellular carcinoma on dynamic contrast-enhanced MRI based on deep learning

Objective. Precise hepatocellular carcinoma (HCC) detection is crucial for clinical management. While studies focus on computed tomography-based automatic algorithms, there is a rareness of research on automatic detection based on dynamic contrast enhanced (DCE) magnetic resonance imaging. This study is to develop an automatic detection and segmentation deep learning model for HCC using DCE. Approach: DCE images acquired from 2016 to 2021 were retrospectively collected. Then, 382 patients (301 male; 81 female) with 466 lesions pathologically confirmed were included and divided into an 80% training-validation set and a 20% independent test set. For external validation, 51 patients (42 male; 9 female) in another hospital from 2018 to 2021 were included. The U-net architecture was modified to accommodate multi-phasic DCE input. The model was trained with the training-validation set using five-fold cross-validation, and furtherly evaluated with the independent test set using comprehensive metrics for segmentation and detection performance. The proposed automatic segmentation model consisted of five main steps: phase registration, automatic liver region extraction using a pre-trained model, automatic HCC lesion segmentation using the multi-phasic deep learning model, ensemble of five-fold predictions, and post-processing using connected component analysis to enhance the performance to refine predictions and eliminate false positives. Main results. The proposed model achieved a mean dice similarity coefficient (DSC) of 0.81 ± 0.11, a sensitivity of 94.41 ± 15.50%, a precision of 94.19 ± 17.32%, and 0.14 ± 0.48 false positive lesions per patient in the independent test set. The model detected 88% (80/91) HCC lesions in the condition of DSC > 0.5, and the DSC per tumor was 0.80 ± 0.13. In the external set, the model detected 92% (58/62) lesions with 0.12 ± 0.33 false positives per patient, and the DSC per tumor was 0.75 ± 0.10. Significance. This study developed an automatic detection and segmentation deep learning model for HCC using DCE, which yielded promising post-processed results in accurately identifying and delineating HCC lesions.

Fully Automated Identification of Lymph Node Metastases and Lymphovascular Invasion in Endometrial Cancer From Multi-Parametric MRI by Deep Learning.

Early and accurate identification of lymphatic node metastasis (LNM) and lymphatic vascular space invasion (LVSI) for endometrial cancer (EC) patients is important for treatment design, but difficult on multi-parametric MRI (mpMRI) images. To develop a deep learning (DL) model to simultaneously identify of LNM and LVSI of EC from mpMRI images. Retrospective. Six hundred twenty-one patients with histologically proven EC from two institutions, including 111 LNM-positive and 168 LVSI-positive, divided into training, internal, and external test cohorts of 398, 169, and 54 patients, respectively. T2-weighted imaging (T2WI), contrast-enhanced T1WI (CE-T1WI), and diffusion-weighted imaging (DWI) were scanned with turbo spin-echo, gradient-echo, and two-dimensional echo-planar sequences, using either a 1.5 T or 3 T system. EC lesions were manually delineated on T2WI by two radiologists and used to train an nnU-Net model for automatic segmentation. A multi-task DL model was developed to simultaneously identify LNM and LVSI positive status using the segmented EC lesion regions and T2WI, CE-T1WI, and DWI images as inputs. The performance of the model for LNM-positive diagnosis was compared with those of three radiologists in the external test cohort. Dice similarity coefficient (DSC) was used to evaluate segmentation results. Receiver Operating Characteristic (ROC) analysis was used to assess the performance of LNM and LVSI status identification. P value <0.05 was considered significant. EC lesion segmentation model achieved mean DSC values of 0.700 ± 0.25 and 0.693 ± 0.21 in the internal and external test cohorts, respectively. For LNM positive/LVSI positive identification, the proposed model achieved AUC values of 0.895/0.848, 0.806/0.795, and 0.804/0.728 in the training, internal, and external test cohorts, respectively, and better than those of three radiologists (AUC = 0.770/0.648/0.674). The proposed model has potential to help clinicians to identify LNM and LVSI status of EC patients and improve treatment planning. 3 TECHNICAL EFFICACY: Stage 2.