Interactive contouring through contextual deep learning.

BackgroundNasopharyngeal carcinoma (NPC), an endemic malignancy in Southeast Asia and southern China, necessitates precise delineation of the gross tumor volume (GTV) for radiotherapy to optimize patient prognosis. Positron emission tomography (PET) imaging offers valuable metabolic insights to guide radiotherapy planning but is hampered by manual GTV segmentation, which is labor-intensive and prone to inter- and intra-observer variability. Conventional segmentation methods are inherently limited, and comparative evaluations of deep learning (DL) frameworks for PET-based primary NPC lesion segmentation remain scarce. Thus, this study aimed to optimize the delineation of primary NPC radiotherapy target volumes by leveraging PET-image-based DL segmentation technology, with the goal of enhancing both the precision and efficiency of lesion delineation.MethodsResearchers retrospectively collected PET imaging data from 212 NPC patients at the Affiliated Hospital of Guangdong Medical University. The patients were randomly divided into training (170 patients) and testing (42 patients) datasets. A radiation oncologist and a nuclear medicine physician collaboratively delineated the cancer lesion boundaries through consensus. After data preprocessing, three DL models (Res-Unet, Nn-Unet, and Nn-Former) were used to automatically segment the lesions. Training was based on the training dataset. During evaluation, the models’ segmentation performance was comprehensively assessed using Dice similarity coefficient (DSC) and Hausdorff distance (HD). A visual analysis of the results was also conducted to intuitively understand the models’ segmentation capabilities.ResultsIn the training set, Nn-Unet achieved the highest DSC of 0.869, whereas in the testing set, its DSC was 0.833. The DSC for Nn-Former hovered around 0.8 in both the training and testing sets. In contrast, Res-Unet demonstrated the lowest DSC values among the three models, particularly in the testing set (DSC =0.794). Statistical analysis revealed that the DSC of Nn-Unet was significantly higher than that of both Res-Unet and Nn-Former in the training set (P<0.01 for both comparisons). Regarding the 95% HD and average surface distance (ASD) values, Nn-Unet outperformed the other two models in both the training and testing sets. However, Res-Unet exhibited a much higher HD value in the testing set compared to the training set. The Loss curves for all models gradually decreased as training progressed, indicating that the models were learning relevant features of NPC lesions. The final Loss value for Res-Unet was approximately 0.7, whereas Nn-Unet and Nn-Former had final Loss values below 0.8.ConclusionsThe PET-image-based automatic segmentation model for the primary tumor of NPC established in this study demonstrates the clinical potential of Nn-Unet in primary lesion of NPC segmentation tasks.

Manual segmentation in lumbar paraspinal muscle studies is time-consuming, subject to variability, and limits comprehensive evaluation. To solve the problems, we aim to develop a deep learning algorithm for the automatic segmentation and multi-parameter quantification of eight paraspinal muscles in lumbar CT. We collected CT images covering the range from the superior endplate of the L1 vertebra to the inferior endplate of the S1 vertebra. The dataset was partitioned into training, validation, and test sets in a ratio of 7:1:2. Six convolutional neural networks were employed to automatically segment the bilateral psoas major, quadratus lumborum, erector spinae, and multifidus across all slices. Model performance was evaluated using the Dice Similarity Coefficient (DSC), Hausdorff Distance (HD), and mean Intersection over Union (mIoU). Muscle parameters including cross-sectional area, muscle volume, fat infiltration, CT density, and paraspinal muscle index were calculated. The Intraclass Correlation Coefficient [ICC(3,1)] was used to assess agreement between model's segmentations and manual ground truth. 100 lumbar CT scans (age, 55.02 ± 16.2; 62 female) were used for model development. The mean DSC across all six models ranged from 0.754 to 0.903. TransUNet demonstrated the best overall (DSC, 0.903; 95% CI: 0.902, 0.913) (HD, 9.934mm; 95% CI: 9.815, 11.902) (mIoU, 0.841; 95% CI: 0.839, 0.850). For individual muscles segmented by TransUNet, the highest DSC was 0.936 (95% CI: 0.934-0.938) for the right multifidus. The mean ICC for all parameters was 0.931. The highest ICC value was 0.999 (p < 0.001). The deep learning tool enables automatic and accurate segmentation of eight lumbar paraspinal muscles. Muscle parameter measurements derived from model's segmentations and manual ground truth exhibit high agreement. Accurate segmentation and quantification of paraspinal muscle parameters facilitates large-scale epidemiological studies of paraspinal muscles and spine-related diseases, overcoming the limitations in manual methods.

SummaryAccurate prediction of the solubility of chemical substances in solvents remains a challenge. The sparsity of high-quality solubility data is recognized as the biggest hurdle in the development of robust data-driven methods for practical use. Nonetheless, the effects of the quality and quantity of data on aqueous solubility predictions have not yet been scrutinized. In this study, the roles of the size and the quality of data sets on the performances of the solubility prediction models are unraveled, and the concepts of actual and observed performances are introduced. In an effort to curtail the gap between actual and observed performances, a quality-oriented data selection method, which evaluates the quality of data and extracts the most accurate part of it through statistical validation, is designed. Applying this method on the largest publicly available solubility database and using a consensus machine learning approach, a top-performing solubility prediction model is achieved.

The vessel-plaque interface detection and lesion length of coronary artery disease are still challenging using the deep learning with intravascular ultrasound (IVUS). We used AI-based image segmentation technology to develop a deep learning model for classifying frames with IVUS-derived target lesion. Methods. A total of 3,782 target lesion frames from 28 acute coronary syndrome(ACS) patients were randomized into training and test sets in a 2:1 ratio. The training algorithm was based on the measurement of the core lab. A segmentation model using non-local blocks was used to increase the similarity between pixels of the same class in a U-Net-based model. The edge detection was performed on the region of interest, and the detected edge was used as an aid in deriving the final segmentation result. And the morphological guided measurements were used to provide the information about the vessel area, lumen area, plaque burden and lesion length. Quantitative evaluation was conducted on four scales: Jaccard Measure (JM), Dice Similarity Coefficient (DSC), Hausdorff Distance (HD), and Percentage of Area Difference (PAD). And we compared the diagnostic accuracy to detect the vessel-plaque interface of the target lesion. Results. In the training sample (1,256 IVUS frames of target lesion), and the test samples (1,072 IVUS frames of target lesion) were used for the proposed model. The external elastic membrane area (EEM), lumen area, plaque area and thrombus were assessed with the proposed algorithm. The measurement of EEM was DSC 0.99, JM 0.98, HD 0.06, and PAD 0.01. And the lumen also shows excellent performance with DSC 0.98, JM 0.97, HD 0.08, and PAD 0.01. In addition, the proposed model was compared with ground-truth for Thrombus, and showed high values of 0.88 for DSC, 0.83 for JM, 0.21 for HD, and 0.09 for PAD. Conclusion. An advanced algorithm can accurately detect the vessel-plaque interface and estimated stent length of coronary artery disease with high reproducibility. F

Segmentation of cardiac structures is essential for cardiac function evaluation using cine magnetic resonance imaging (MRI). Deep learning can be used to segment cardiac structures in cine cardiac MRI with high accuracy, but this approach requires fully annotated datasets for training, which are difficult to obtain. Semi-supervised segmentation methods provide a way to alleviate the burden of manual labeling by using labeled and unlabeled data for training. However, these methods generally provide suboptimal segmentationaccuracies. To develop a semi-supervised method that utilizes relatively small training datasets and under-annotations for improved cine cardiac MRIsegmentation. The proposed approach consists of deformable registration, fully and weakly supervised segmentation, and a temporal attention perceiver (TAP). The registration module was trained to warp labeled frames to generate pseudo labels for unlabeled frames. The warped labeled images were used to train the fully supervised segmentation network. The unlabeled images and the pseudo label were used to train the weakly supervised segmentation model, and the segmentation prediction was compared with the input pseudo label as an auxiliary loss to the registration module. The TAP module was employed to generate optimized features for the warped labeled and the original unlabeled images both paired with the original labeled image. Consistency between the resulting features was enforced to refine cross-instance feature alignment to facilitate the registration. One hundred, twenty, and ten subjects from the Automatic Cardiac Diagnosis Challenge (ACDC) and seventy-five, thirty, and fifteen cases from the Multi-Vendor & Multi-Disease (M&Ms) Cardiac Image Segmentation Challenge were used for training, each with random end-systolic (ES)/end-diastolic (ED) frames labeled. The optimized models were used to segment the remaining 50 ACDC and 50 M&Ms subjects. The proposed approach was compared with several commonly used semi-supervised segmentation methods in terms of Dice-similarity-coefficients (DSC), average-symmetric-surface-distance (ASSD), and Hausdorff-distance (HD) for left (LV) and right (RV) ventricular cavity and myocardium (Myo). A Unet trained on the same subjects each with both frames labeled was used as an upper bound (Unet_UB). Using 100 ACDC training subjects, our approach yielded DSC = 0.910 0.063, ASSD = 1.37 0.63 mm, and HD = 6.38 2.99 mm for RV, DSC = 0.894 0.024, ASSD = 1.20 1.12 mm, and HD = 4.67 3.22 mm for Myo, and DSC = 0.934 0.056, ASSD = 1.25 1.63 mm, and HD = 3.97 5.76 mm for LV. A bidirectional copy-paste (BCP) method performed the best among the comparative methods and generated DSC = 0.902 0.060, ASSD = 1.45 0.60 mm, and HD = 7.50 3.20 mm for RV, DSC = 0.885 0.030, ASSD = 1.28 0.80 mm, and HD = 5.80 2.80 mm for Myo, and DSC = 0.920 0.068, ASSD = 1.15 0.40 mm, and HD = 4.20 3.30 mm for LV. For Unet_UB, these were 0.905 0.068, 1.48 0.61 mm, and 6.35 2.85 mm for RV, 0.895 0.030, 1.05 0.45 mm, and 4.40 3.09 mm for Myo, and 0.941 0.044, 1.02 0.34 mm, and 3.17 1.63 mm for LV. Similar trends were observed when using 75 M&Ms training subjects. For all the experiments, our approach outperformed BCP in general and yielded segmentation accuracies comparable to Unet_UB. The proposed approach outperformed several commonly used semi-supervised segmentation methods and yielded segmentation accuracies on par with fully supervised Unet using various relatively small datasets and under annotations fortraining.

Deep Learning Method for Rapid Simultaneous Multistructure Temporal Bone Segmentation.

Implementation of automation routines leveraging deep learning (DL) methods has been a growing topic of interest. The focus of this work is on the Gamma Knife (GK) workflow, specifically the approximation of dose distributions from GK plan parameters and the prediction of dose distributions using DL. These works contribute towards the larger goal of treatment plan prediction and are seen as intermediate steps towards that end. The approximation of dose distributions was motivated by the closed nature of the GK system, which causes complications with the evaluation of the predictions made by the DL models. The approximation utilizes a superposition method, accumulating dose from a basis set of dose distributions which correspond to specific sector activations of the GK system. The approximation method was compared to the original dose distributions calculated by the GK treatment planning system (TPS), Leksell GammaPlan, for 30 patients using the similarity metrics of Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD) as well as the plan quality metrics of coverage, selectivity, and gradient index. In the range of clinical planning significance, the metrics reported strong agreement to the original calculation (DSC > 0.85, HD < 2.5mm, coverage difference = 0.014 ± 0.035, selectivity difference = 0.008 ± 0.037, and gradient index difference = 0.118 ± 0.192). The second focus of this work was to improve the performance of DL models used to predict dose distributions from target contour information. This was accomplished by applying transforms to the binary target contour volumes, resulting in alternate model input channels that encode the distance of each voxel to the boundary of the nearest contour, both within and outside of the contours. Training and evaluation were performed with a standard UNet and HD-UNet architecture for large and small targets separately. DSC and Predictions resulting from these encodings resulted in significantly better predictions (p < 0.05) than their binary target counterparts at all (excluding 90% isodose for the HD-Unet) threshold levels for small targets. The large target trainings were significantly (p < 0.05) improved in the mid-to-high dose ranges.

We propose a fully-automatic deep learning-based algorithm for segmentation of ocular structures and microbial keratitis (MK) biomarkers on slit-lamp photography (SLP) images. The dataset consisted of SLP images from 133 eyes with manual annotations by a physician, P1. A modified region-based convolutional neural network, SLIT-Net, was developed and trained using P1's annotations to identify and segment four pathological regions of interest (ROIs) on diffuse white light images (stromal infiltrate (SI), hypopyon, white blood cell (WBC) border, corneal edema border), one pathological ROI on diffuse blue light images (epithelial defect (ED)), and two non-pathological ROIs on all images (corneal limbus, light reflexes). To assess inter-reader variability, 75 eyes were manually annotated for pathological ROIs by a second physician, P2. Performance was evaluated using the Dice similarity coefficient (DSC) and Hausdorff distance (HD). Using seven-fold cross-validation, the DSC of the algorithm (as compared to P1) for all ROIs was good (range: 0.62-0.95) on all 133 eyes. For the subset of 75 eyes with manual annotations by P2, the DSC for pathological ROIs ranged from 0.69-0.85 (SLIT-Net) vs. 0.37-0.92 (P2). DSCs for SLIT-Net were not significantly different than P2 for segmenting hypopyons (p > 0.05) and higher than P2 for WBCs (p < 0.001) and edema (p < 0.001). DSCs were higher for P2 for segmenting SIs (p < 0.001) and EDs (p < 0.001). HDs were lower for P2 for segmenting SIs (p = 0.005) and EDs (p < 0.001) and not significantly different for hypopyons (p > 0.05), WBCs (p > 0.05), and edema (p > 0.05). This prototype fully-automatic algorithm to segment MK biomarkers on SLP images performed to expectations on an exploratory dataset and holds promise for quantification of corneal physiology and pathology.

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.

Application of contrast-enhanced CT-driven multimodal machine learning models for pulmonary metastasis prediction in head and neck adenoid cystic carcinoma.

The objective of this study is to investigate the value of radiomics features and deep learning features based on positron emission tomography/computed tomography (PET/CT) in predicting perineural invasion (PNI) in rectal cancer. We retrospectively collected 120 rectal cancer (56 PNI-positive patients 64 PNI-negative patients) patients with preoperative 18F-FDG PET/CT examination and randomly divided them into training and validation sets at a 7:3 ratio. We also collected 31 rectal cancer patients from two other hospitals as an independent external validation set. χ2 test and binary logistic regression were used to analyze PET metabolic parameters. PET/CT images were utilized to extract radiomics features and deep learning features. The Mann-Whitney U test and LASSO were employed to select valuable features. Metabolic parameter, radiomics, deep learning and combined models were constructed. ROC curves were generated to evaluate the performance of models. The results indicate that metabolic tumor volume (MTV) is correlated with PNI (P = 0.001). In the training set and validation set, the AUC values of the metabolic parameter model were 0.673 (95%CI: 0.572-0.773), 0.748 (95%CI: 0.599-0.896). We selected 16 radiomics features and 17 deep learning features as valuable factors for predicting PNI. The AUC values of radiomics model and deep learning model were 0.768 (95%CI: 0.667-0.868) and 0.860 (95%CI: 0.780-0.940) in the training set. And the AUC values in the validation set were 0.803 (95%CI: 0.656-0.950) and 0.854 (95% CI 0.721-0.987). Finally, the combined model exhibited AUCs of 0.893 (95%CI: 0.825-0.961) in the training set and 0.883 (95%CI: 0.775-0.990) in the validation set. In the external validation set, the combined model achieved an AUC of 0.829 (95% CI: 0.674-0.984), outperforming each individual model. The decision curve analysis of these models indicated that using the combined model to guide treatment provided a substantial net benefit. This combined model established by integrating PET metabolic parameters, radiomics features, and deep learning features can accurately predict the PNI in rectal cancer.

PCOSt: A non-invasive and cost-effective screening tool for polycystic ovary syndrome

The reaserch of artificial intelligence (AI) model for predicting spinal refracture is limited to bone mineral density, X-ray and some conventional laboratory indicators, which has its own limitations. Besides, it lacks specific indicators related to osteoporosis and imaging factors that can better reflect bone quality, such as computed tomography (CT). To construct a novel predicting model based on bone turn-over markers and CT to identify patients who were more inclined to suffer spine refracture. CT images and clinical information of 383 patients (training set = 240 cases of osteoporotic vertebral compression fractures (OVCF), validation set = 63, test set = 80) were retrospectively collected from January 2015 to October 2022 at three medical centers. The U-net model was adopted to automatically segment ROI. Three-dimensional (3D) cropping of all spine regions was used to achieve the final ROI regions including 3D_Full and 3D_RoiOnly. We used the Densenet 121-3D model to model the cropped region and simultaneously build a T-NIPT prediction model. Diagnostics of deep learning models were assessed by constructing ROC curves. We generated calibration curves to assess the calibration performance. Additionally, decision curve analysis (DCA) was used to assess the clinical utility of the predictive models. The performance of the test model is comparable to its performance on the training set (dice coefficients of 0.798, an mIOU of 0.755, an SA of 0.767, and an OS of 0.017). Univariable and multivariable analysis indicate that T_P1NT was an independent risk factor for refracture. The performance of predicting refractures in different ROI regions showed that 3D_Full model exhibits the highest calibration performance, with a Hosmer-Lemeshow goodness-of-fit (HL) test statistic exceeding 0.05. The analysis of the training and test sets showed that the 3D_Full model, which integrates clinical and deep learning results, demonstrated superior performance with significant improvement (p-value < 0.05) compared to using clinical features independently or using only 3D_RoiOnly. T_P1NT was an independent risk factor of refracture. Our 3D-FULL model showed better performance in predicting high-risk population of spine refracture than other models and junior doctors do. This model can be applicable to real-world translation due to its automatic segmentation and detection.

MRI represents the most established liver iron content (LIC) evaluation approach by estimation of liver T2* value, but it is dependent on the choice of the measurement region and the software used for image analysis. To develop a deep-learning method for unsupervised classification of LIC from magnitude T2* multiecho MR images. Retrospective. A total of 1069 thalassemia major patients enrolled in the core laboratory of the Myocardial Iron Overload in Thalassemia (MIOT) network, which were included in the training (80%) and test (20%) sets. Twenty patients from different MRI vendors included in the external test set. A5 T, T2* multiecho magnitude images. Four deep-learning convolutional neural networks (HippoNet-2D, HippoNet-3D, HippoNet-LSTM, and an ensemble network HippoNet-Ensemble) were used to achieve unsupervised staging of LIC using five classes (normal, borderline, middle, moderate, severe). The training set was employed to construct the deep-learning model. The performance of the LIC staging model was evaluated in the test set and in the external test set. The model's performances were assessed by evaluating the accuracy, sensitivity, and specificity with respect to the ground truth labels obtained by T2* measurements and by comparison with operator-induced variability originating from different region of interest (ROI) placements. The network's performances were evaluated by single-class accuracy, specificity, and sensitivity and compared by one-way repeated measures analysis of variance (ANOVA) and one-way ANOVA. HippoNet-Ensemble reached an accuracy significantly higher than the other networks, and a sensitivity and specificity higher than HippoNet-LSTM. Accuracy, sensitivity, and specificity values for the LIC stages were: normal: 0.96/0.93/0.97, borderline: 0.95/0.85/0.98, mild: 0.96/0.88/0.98, moderate: 0.95/0.89/0.97, severe: 0.97/0.95/0.98. Correctly staging of cases was in the range of 85%-95%, depending on the LIC class. Multiclass accuracy was 0.90 against 0.92 for the interobserver variability. The proposed HippoNet-Ensemble network can perform unsupervised LIC staging and achieves good prognostic performance. 4 TECHNICAL EFFICACY: Stage 2.

Cardiac structure contouring is a time consuming and tedious manual activity used for radiotherapeutic dose toxicity planning. We developed an automatic cardiac structure segmentation pipeline for use in low-dose non-contrast planning CT based on deep learning algorithms for small datasets. Fifty CT scans were retrospectively selected and the whole heart, ventricles and atria were contoured. A two stage deep learning pipeline was trained on 41 non contrast planning CTs, tuned with 3 CT scans and validated on 6 CT scans. In the first stage, An InceptionResNetV2 network was used to identify the slices that contained cardiac structures. The second stage consisted of three deep learning models trained on the images containing cardiac structures to segment the structures. The three deep learning models predicted the segmentations/contours on axial, coronal and sagittal images and are combined to create the final prediction. The final accuracy of the pipeline was quantified on 6 volumes by calculating the Dice similarity coefficient (DC), 95% Hausdorff distance (95% HD) and volume ratios between predicted and ground truth volumes. Median DC and 95% HD of 0.96, 0.88, 0.92, 0.80 and 0.82, and 1.86, 2.98, 2.02, 6.16 and 6.46 were achieved for the whole heart, right and left ventricle, and right and left atria respectively. The median differences in volume were -4, -1, + 5, -16 and -20% for the whole heart, right and left ventricle, and right and left atria respectively. The automatic contouring pipeline achieves good results for whole heart and ventricles. Robust automatic contouring with deep learning methods seems viable for local centers with small datasets.