Highest Dice Similarity Coefficient Research Articles

An end-to-end multi-scale airway segmentation framework based on pulmonary CT image

Objective. Automatic and accurate airway segmentation is necessary for lung disease diagnosis. The complex tree-like structures leads to gaps in the different generations of the airway tree, and thus airway segmentation is also considered to be a multi-scale problem. In recent years, convolutional neural networks have facilitated the development of medical image segmentation. In particular, 2D CNNs and 3D CNNs can extract different scale features. Hence, we propose a two-stage and 2D + 3D framework for multi-scale airway tree segmentation. Approach. In stage 1, we use a 2D full airway SegNet(2D FA-SegNet) to segment the complete airway tree. Multi-scale atros spatial pyramid and Atros Residual Skip connection modules are inserted to extract different scales feature. We designed a hard sample selection strategy to increase the proportion of intrapulmonary airway samples in stage 2. 3D airway RefineNet (3D ARNet) as stage 2 takes the results of stage 1 as a priori information. Spatial information extracted by 3D convolutional kernel compensates for the loss of in 2D FA-SegNet. Furthermore, we added false positive losses and false negative losses to improve the segmentation performance of airway branches within the lungs. Main results. We performed data enhancement on the publicly available dataset of ISICDM 2020 Challenge 3, and on which evaluated our method. Comprehensive experiments show that the proposed method has the highest dice similarity coefficient (DSC) of 0.931, and IoU of 0.871 for the whole airway tree and DSC of 0.699, and IoU of 0.543 for the intrapulmonary bronchi tree. In addition, 3D ARNet proposed in this paper cascaded with other state-of-the-art methods to increase detected tree length rate by up to 46.33% and detected tree branch rate by up to 42.97%. Significance. The quantitative and qualitative evaluation results show that our proposed method performs well in segmenting the airway at different scales.

Hybrid deep learning and optimal graph search method for optical coherence tomography layer segmentation in diseases affecting the optic nerve

Accurate segmentation of retinal layers in optical coherence tomography (OCT) images is critical for assessing diseases that affect the optic nerve, but existing automated algorithms often fail when pathology causes irregular layer topology, such as extreme thinning of the ganglion cell-inner plexiform layer (GCIPL). Deep LOGISMOS, a hybrid approach that combines the strengths of deep learning and 3D graph search to overcome their limitations, was developed to improve the accuracy, robustness and generalizability of retinal layer segmentation. The method was trained on 124 OCT volumes from both eyes of 31 non-arteritic anterior ischemic optic neuropathy (NAION) patients and tested on three cross-sectional datasets with available reference tracings: Test-NAION (40 volumes from both eyes of 20 NAION subjects), Test-G (29 volumes from 29 glaucoma subjects/eyes), and Test-JHU (35 volumes from 21 multiple sclerosis and 14 control subjects/eyes) and one longitudinal dataset without reference tracings: Test-G-L (155 volumes from 15 glaucoma patients/eyes). In the three test datasets with reference tracings (Test-NAION, Test-G, and Test-JHU), Deep LOGISMOS achieved very high Dice similarity coefficients (%) on GCIPL: 89.97±3.59, 90.63±2.56, and 94.06±1.76, respectively. In the same context, Deep LOGISMOS outperformed the Iowa reference algorithms by improving the Dice score by 17.5, 5.4, and 7.5, and also surpassed the deep learning framework nnU-Net with improvements of 4.4, 3.7, and 1.0. For the 15 severe glaucoma eyes with marked GCIPL thinning (Test-G-L), it demonstrated reliable regional GCIPL thickness measurement over five years. The proposed Deep LOGISMOS approach has potential to enhance precise quantification of retinal structures, aiding diagnosis and treatment management of optic nerve diseases.

Automated segmentation of lesions and organs at risk on [68Ga]Ga-PSMA-11 PET/CT images using self-supervised learning with Swin UNETR

BackgroundProstate-specific membrane antigen (PSMA) PET/CT imaging is widely used for quantitative image analysis, especially in radioligand therapy (RLT) for metastatic castration-resistant prostate cancer (mCRPC). Unknown features influencing PSMA biodistribution can be explored by analyzing segmented organs at risk (OAR) and lesions. Manual segmentation is time-consuming and labor-intensive, so automated segmentation methods are desirable. Training deep-learning segmentation models is challenging due to the scarcity of high-quality annotated images. Addressing this, we developed shifted windows UNEt TRansformers (Swin UNETR) for fully automated segmentation. Within a self-supervised framework, the model’s encoder was pre-trained on unlabeled data. The entire model was fine-tuned, including its decoder, using labeled data.MethodsIn this work, 752 whole-body [68Ga]Ga-PSMA-11 PET/CT images were collected from two centers. For self-supervised model pre-training, 652 unlabeled images were employed. The remaining 100 images were manually labeled for supervised training. In the supervised training phase, 5-fold cross-validation was used with 64 images for model training and 16 for validation, from one center. For testing, 20 hold-out images, evenly distributed between two centers, were used. Image segmentation and quantification metrics were evaluated on the test set compared to the ground-truth segmentation conducted by a nuclear medicine physician.ResultsThe model generates high-quality OARs and lesion segmentation in lesion-positive cases, including mCRPC. The results show that self-supervised pre-training significantly improved the average dice similarity coefficient (DSC) for all classes by about 3%. Compared to nnU-Net, a well-established model in medical image segmentation, our approach outperformed with a 5% higher DSC. This improvement was attributed to our model’s combined use of self-supervised pre-training and supervised fine-tuning, specifically when applied to PET/CT input. Our best model had the lowest DSC for lesions at 0.68 and the highest for liver at 0.95.ConclusionsWe developed a state-of-the-art neural network using self-supervised pre-training on whole-body [68Ga]Ga-PSMA-11 PET/CT images, followed by fine-tuning on a limited set of annotated images. The model generates high-quality OARs and lesion segmentation for PSMA image analysis. The generalizable model holds potential for various clinical applications, including enhanced RLT and patient-specific internal dosimetry.

Deep Learning-Based Lumen and Vessel Segmentation of Intravascular Ultrasound Images in Coronary Artery Disease.

Intravascular ultrasound (IVUS) evaluation of coronary artery morphology is based on the lumen and vessel segmentation. This study aimed to develop an automatic segmentation algorithm and validate the performances for measuring quantitative IVUS parameters. A total of 1,063 patients were randomly assigned, with a ratio of 4:1 to the training and test sets. The independent data set of 111 IVUS pullbacks was obtained to assess the vessel-level performance. The lumen and external elastic membrane (EEM) boundaries were labeled manually in every IVUS frame with a 0.2-mm interval. The Efficient-UNet was utilized for the automatic segmentation of IVUS images. At the frame-level, Efficient-UNet showed a high dice similarity coefficient (DSC, 0.93±0.05) and Jaccard index (JI, 0.87±0.08) for lumen segmentation, and demonstrated a high DSC (0.97±0.03) and JI (0.94±0.04) for EEM segmentation. At the vessel-level, there were close correlations between model-derived vs. experts-measured IVUS parameters; minimal lumen image area (r=0.92), EEM area (r=0.88), lumen volume (r=0.99) and plaque volume (r=0.95). The agreement between model-derived vs. expert-measured minimal lumen area was similarly excellent compared to the experts' agreement. The model-based lumen and EEM segmentation for a 20-mm lesion segment required 13.2 seconds, whereas manual segmentation with a 0.2-mm interval by an expert took 187.5 minutes on average. The deep learning models can accurately and quickly delineate vascular geometry. The artificial intelligence-based methodology may support clinicians' decision-making by real-time application in the catheterization laboratory.

Feasibility study of adaptive radiotherapy with Ethos for breast cancer.

The aim of this study was to assess the feasibility of online adaptive radiotherapy with Ethos for breast cancer. This retrospective study included 20 breast cancer patients previously treated with TrueBeam. All had undergone breast surgery for different indications (right/left, lumpectomy/mastectomy) and were evenly divided between these four cases, with five extended cone beam computed tomography (CBCT) scans per patient. The dataset was used in an Ethos emulator to test the full adaptive workflow. The contours generated by artificial intelligence (AI) for the influencers (left and right breasts and lungs, heart) and elastic or rigid propagation for the target volumes (internal mammary chain (IMC) and clavicular lymph nodes (CLNs)) were compared to the initial contours delineated by the physician using two metrics: Dice similarity coefficient (DICE) and Hausdorff 95% distance (HD95). The repeatability of influencer generation was investigated. The times taken by the emulator to generate contours, optimize plans, and calculate doses were recorded. The quality of the scheduled and adapted plans generated by Ethos was assessed using planning target volume (PTV) coverage, homogeneity indices (HIs), and doses to organs at risk (OARs) via dose-volume histogram (DVH) metrics. Quality assurance (QA) of the treatment plans was performed using an independent portal dosimetry tool (EpiQA) and gamma index. On average, the DICE for the influencers was greater than 0.9. Contours resulting from rigid propagation had a higher DICE and a lower HD95 than those resulting from elastic deformation but remained below the values obtained for the influencers: DICE values were 0.79 ± 0.11 and 0.46 ± 0.17 for the CLN and IMC, respectively. Regarding the repeatability of the influencer segmentation, the DICE was close to 1, and the mean HD95 was strictly less than 0.15mm. The mean time was 73 ± 4 s for contour generation per AI and 80 ± 9 s for propagations. The average time was 53 ± 3 s for dose calculation and 125 ± 9 s for plan optimization. A dosimetric comparison of scheduled and adapted plans showed a significant difference in PTV coverage: dose received by 95% of the volume (D95%) values were higher and closer to the prescribed doses for adapted plans. Doses to organs at risk were similar. The average gamma index for quality assurance of adapted plans was 99.93 ± 0.38 for a 3%/3mm criterion. This study comprehensively evaluated the Ethos® adaptive workflow for breast cancer and its potential technical limitations. Although the results demonstrated the high accuracy of AI segmentation and the superiority of adapted plans in terms of target volume coverage, a medical assessment is still required.

Impact on Reconstruction SPECT Image Distortion by Iteration and Subset Number

Iterative reconstruction on Single Photon Emission Computed Tomography (SPECT) imaging has recently been commercially available, resulting in reconstructed images with lower noise levels and better spatial resolution. The reconstruction variables employed and chosen under these suggestions include post-filtering, as well as several iterations and subset numbers. However, most of the number of iterations and subsets will give distortion to the segmentation image on a volume to be segmented. For this purpose, this paper will compare the effect of distortion on the dice similarity coefficient of volume segmentation for the iterative reconstruction setting by default as practice. The best outcome with the smallest root-mean-square deviation, highest percentage contrast values, and highest dice similarity coefficient) for all iterations and subsets is in iteration 6 subset 16. Here, the root-mean-square deviation is 8.06. Moreover, the four most enormous spheres can be calculated from the six spheres filled. The best contrast percentage is in subset 32, but the background variability for that subset is exceptionally high, which is 64.5. With post-filtering implementation, the scatter-corrected slices show the most enhanced image resolution and contrast when using advanced Astonish reconstruction algorithms. The user must decide if noise reduction via post-filtering or increased image resolution is more suitable for a certain imaging process.

Logarithmic total variation regularization via preconditioned conjugate gradient method for sparse reconstruction of bioluminescence tomography

Background and objectiveBioluminescence Tomography (BLT) is a powerful optical molecular imaging technique that enables the noninvasive investigation of dynamic biological phenomena. It aims to reconstruct the three-dimensional spatial distribution of bioluminescent sources from optical measurements collected on the surface of the imaged object. However, BLT reconstruction is a challenging ill-posed problem due to the scattering effect of light and the limitations in detecting surface photons, which makes it difficult for existing methods to achieve satisfactory reconstruction results. In this study, we propose a novel method for sparse reconstruction of BLT based on a preconditioned conjugate gradient with logarithmic total variation regularization (PCG-logTV). MethodThis PCG-logTV method incorporates the sparsity of overlapping groups and enhances the sparse structure of these groups using logarithmic functions, which can preserve edge features and achieve more stable reconstruction results in BLT. To accelerate the convergence of the algorithm solution, we use the preconditioned conjugate gradient iteration method on the objective function and obtain the reconstruction results. We demonstrate the performance of our proposed method through numerical simulations and in vivo experiment. Results and conclusionsThe results show that the PCG-logTV method obtains the most accurate reconstruction results, and the minimum position error (LE) is 0.254mm, which is 26%, 31% and 34% of the FISTA (0.961), IVTCG (0.81) and L1-TV (0.739) methods, and the root mean square error (RMSE) and relative intensity error (RIE) are the smallest, indicating that it is closest to the real light source. In addition, compared with the other three methods, the PCG-logTV method also has the highest DICE similarity coefficient, which is 0.928, which means that this method can effectively reconstruct the three-dimensional spatial distribution of bioluminescent light sources, has higher resolution and robustness, and is beneficial to the preclinical and clinical studies of BLT.

Feasibility of artificial intelligence-driven interfractional monitoring of organ changes by mega-voltage computed tomography in intensity-modulated radiotherapy of prostate cancer.

High-dose radiotherapy (RT) for localized prostate cancer requires careful consideration of target position changes and adjacent organs-at-risk (OARs), such as the rectum and bladder. Therefore, daily monitoring of target position and OAR changes is crucial in minimizing interfractional dosimetric uncertainties. For efficient monitoring of the internal condition of patients, we assessed the feasibility of an auto-segmentation of OARs on the daily acquired images, such as megavoltage computed tomography (MVCT), via a commercial artificial intelligence (AI)-based solution in this study. We collected MVCT images weekly during the entire course of RT for 100 prostate cancer patients treated with the helical TomoTherapy system. Based on the manually contoured body outline, the bladder including prostate area, and rectal balloon regions for the 100 MVCT images, we trained the commercially available fully convolutional (FC)-DenseNet model and tested its auto-contouring performance. Based on the optimally determined hyperparameters, the FC-DenseNet model successfully auto-contoured all regions of interest showing high dice similarity coefficient (DSC) over 0.8 and a small mean surface distance (MSD) within 1.43 mm in reference to the manually contoured data. With this well-trained AI model, we have efficiently monitored the patient's internal condition through six MVCT scans, analyzing DSC, MSD, centroid, and volume differences. We have verified the feasibility of utilizing a commercial AI-based model for auto-segmentation with low-quality daily MVCT images. In the future, we will establish a fast and accurate auto-segmentation and internal organ monitoring system for efficiently determining the time for adaptive replanning.

Joint Segmentation and Differential Diagnosis of Thyroid Nodule in Contrast-Enhanced Ultrasound Images.

Microvascular perfusion can be observed in real time with contrast-enhanced ultrasound (CEUS), which is a novel ultrasound technology for visualizing the dynamic patterns of parenchymal perfusion. Automatic lesion segmentation and differential diagnosis of malignant and benign based on CEUS are crucial but challenging tasks for computer-aided diagnosis of thyroid nodule. To tackle these two formidable challenges concurrently, we provide Trans-CEUS, a spatial-temporal transformer-based CEUS analysis model to finish the joint learning of these two challenging tasks. Specifically, the dynamic swin-transformer encoder and multi-level feature collaborative learning are combined into U-net for achieving accurate segmentation of lesions with ambiguous boundary from CEUS. In addition, variant transformer-based global spatial-temporal fusion is proposed to obtain long-distance enhancement perfusion of dynamic CEUS for promoting differential diagnosis. Empirical results of clinical data showed that our Trans-CEUS model achieved not only a good lesion segmentation result with a high Dice similarity coefficient of 82.41%, but also superior diagnostic accuracy of 86.59%. Conclusion & significance: This research is novel since it is the first to incorporate the transformer into CEUS analysis, and it shows promising results on dynamic CEUS datasets for both segmentation and diagnosis tasks of the thyroid nodule.

A systematic review of automated segmentation of 3D computed-tomography scans for volumetric body composition analysis.

Automated computed tomography (CT) scan segmentation (labelling of pixels according to tissue type) is now possible. This technique is being adapted to achieve three-dimensional (3D) segmentation of CT scans, opposed to single L3-slice alone. This systematic review evaluates feasibility and accuracy of automated segmentation of 3D CT scans for volumetric body composition (BC) analysis, as well as current limitations and pitfalls clinicians and researchers should be aware of. OVID Medline, Embase and grey literature databases up to October 2021 were searched. Original studies investigating automated skeletal muscle, visceral and subcutaneous AT segmentation from CT were included. Seven of the 92 studies met inclusion criteria. Variation existed in expertise and numbers of humans performing ground-truth segmentations used to train algorithms. There was heterogeneity in patient characteristics, pathology and CT phases that segmentation algorithms were developed upon. Reporting of anatomical CT coverage varied, with confusing terminology. Six studies covered volumetric regional slabs rather than the whole body. One study stated the use of whole-body CT, but it was not clear whether this truly meant head-to-fingertip-to-toe. Two studies used conventional computer algorithms. The latter five used deep learning (DL), an artificial intelligence technique where algorithms are similarly organized to brain neuronal pathways. Six of seven reported excellent segmentation performance (Dice similarity coefficients>0.9 per tissue). Internal testing on unseen scans was performed for only four of seven algorithms, whilst only three were tested externally. Trained DL algorithms achieved full CT segmentation in 12 to 75s versus 25min for non-DL techniques. DL enables opportunistic, rapid and automated volumetric BC analysis of CT performed for clinical indications. However, most CT scans do not cover head-to-fingertip-to-toe; further research must validate using common CT regions to estimate true whole-body BC, with direct comparison to single lumbar slice. Due to successes of DL, we expect progressive numbers of algorithms to materialize in addition to the seven discussed in this paper. Researchers and clinicians in the field of BC must therefore be aware of pitfalls. High Dice similarity coefficients do not inform the degree to which BC tissues may be under- or overestimated and nor does it inform on algorithm precision. Consensus is needed to define accuracy and precision standards for ground-truth labelling. Creation of a large international, multicentre common CT dataset with BC ground-truth labels from multiple experts could be a robust solution.

Automated Brain Tumor Detection and Segmentation for Treatment Response Assessment Using Amino Acid PET.

Evaluation of metabolic tumor volume (MTV) changes using amino acid PET has become an important tool for response assessment in brain tumor patients. MTV is usually determined by manual or semiautomatic delineation, which is laborious and may be prone to intra- and interobserver variability. The goal of our study was to develop a method for automated MTV segmentation and to evaluate its performance for response assessment in patients with gliomas. Methods: In total, 699 amino acid PET scans using the tracer O-(2-[18F]fluoroethyl)-l-tyrosine (18F-FET) from 555 brain tumor patients at initial diagnosis or during follow-up were retrospectively evaluated (mainly glioma patients, 76%). 18F-FET PET MTVs were segmented semiautomatically by experienced readers. An artificial neural network (no new U-Net) was configured on 476 scans from 399 patients, and the network performance was evaluated on a test dataset including 223 scans from 156 patients. Surface and volumetric Dice similarity coefficients (DSCs) were used to evaluate segmentation quality. Finally, the network was applied to a recently published 18F-FET PET study on response assessment in glioblastoma patients treated with adjuvant temozolomide chemotherapy for a fully automated response assessment in comparison to an experienced physician. Results: In the test dataset, 92% of lesions with increased uptake (n = 189) and 85% of lesions with iso- or hypometabolic uptake (n = 33) were correctly identified (F1 score, 92%). Single lesions with a contiguous uptake had the highest DSC, followed by lesions with heterogeneous, noncontiguous uptake and multifocal lesions (surface DSC: 0.96, 0.93, and 0.81 respectively; volume DSC: 0.83, 0.77, and 0.67, respectively). Change in MTV, as detected by the automated segmentation, was a significant determinant of disease-free and overall survival, in agreement with the physician's assessment. Conclusion: Our deep learning-based 18F-FET PET segmentation allows reliable, robust, and fully automated evaluation of MTV in brain tumor patients and demonstrates clinical value for automated response assessment.

A semi-supervised learning-based quality evaluation system for digital chest radiographs.

Digital radiography is the most commonly utilized medical imaging technique worldwide, and the quality of radiographs plays a crucial role in accurate disease diagnosis. Therefore, evaluating the quality of radiographs is an essential step in medical examinations. However, manual evaluation can be time-consuming, labor-intensive, and prone to interobserver differences, making it less reliable. To alleviate the workload of radiographic technologists and enhance the efficiency of radiograph quality evaluation, it is crucial to develop rapid and reliable quality evaluation methods and establish a set of quantitative evaluation standards. To address this, we have proposed a quality evaluation system for digital radiographs that utilizes deep learning techniques to achieve fast and precise evaluation. The evaluation of frontal chest radiograph quality involves assessing patient positioning through semantic segmentation and foreign body detection. For lung, scapula, and clavicle segmentation in digital chest radiographs, a residual connection-based convolutional neural network π-ResUNet, was proposed. Criteria for patient positioning evaluation were established based on the segmentation and manual evaluation results. A convolutional neural network, FasterRCNN, was utilized to detect and localize foreign bodies in digital chest radiographs. To enhance the performance of both neural networks, a semi-supervised learning (SSL) strategy was implemented by incorporating a consistency loss that leverages a large number of unlabeled digital radiographs. We also trained the network using the fully supervised learning (FSL) strategy and compared their performance on the test set. The ChestXRay-14 and object-CXR datasets were used throughout the process. By comparing with the manual annotation, the proposed network, trained using the SSL method, achieved a high Dice similarity coefficient (DSC) of 0.96, 0.88, and 0.88 for lung, scapula, and clavicle segmentation, respectively, outperforming the network trained with the FSL method. In addition, for foreign body detection, the proposed SSL method was superior to the FSL method, achieving an AUC (Area under receiver operating characteristic curve, Area under ROC curve) of 0.90 and an FROC (Free-response ROC) of 0.77 on the test dataset. The experimental results show that our proposed system is well-suited for radiograph quality evaluation, with the semi-supervised learning method further improving the network's performance. The proposed method can evaluate the quality of a chest radiograph from two aspects-patient positioning and foreign body detection-within 1 s, offering a promising tool in radiograph quality evaluation.

A Deep Learning-Based Automatic Segmentation and 3D Visualization Technique for Intracranial Hemorrhage Detection Using Computed Tomography Images.

Intracranial hemorrhage (ICH) occurs when blood leaks inside the skull as a result of trauma to the skull or due to medical conditions. ICH usually requires immediate medical and surgical attention because the disease has a high mortality rate, long-term disability potential, and other potentially life-threatening complications. There are a wide range of severity levels, sizes, and morphologies of ICHs, making accurate identification challenging. Hemorrhages that are small are more likely to be missed, particularly in healthcare systems that experience high turnover when it comes to computed tomography (CT) investigations. Although many neuroimaging modalities have been developed, CT remains the standard for diagnosing trauma and hemorrhage (including non-traumatic ones). A CT scan-based diagnosis can provide time-critical, urgent ICH surgery that could save lives because CT scan-based diagnoses can be obtained rapidly. The purpose of this study is to develop a machine-learning algorithm that can detect intracranial hemorrhage based on plain CT images taken from 75 patients. CT images were preprocessed using brain windowing, skull-stripping, and image inversion techniques. Hemorrhage segmentation was performed using multiple pre-trained models on preprocessed CT images. A U-Net model with DenseNet201 pre-trained encoder outperformed other U-Net, U-Net++, and FPN (Feature Pyramid Network) models with the highest Dice similarity coefficient (DSC) and intersection over union (IoU) scores, which were previously used in many other medical applications. We presented a three-dimensional brain model highlighting hemorrhages from ground truth and predicted masks. The volume of hemorrhage was measured volumetrically to determine the size of the hematoma. This study is essential in examining ICH for diagnostic purposes in clinical practice by comparing the predicted 3D model with the ground truth.

Deep learning-based segmentation of dental implants on cone-beam computed tomography images: A validation study

ObjectivesTo train and validate a cloud-based convolutional neural network (CNN) model for automated segmentation (AS) of dental implant and attached prosthetic crown on cone-beam computed tomography (CBCT) images. MethodsA total dataset of 280 maxillomandibular jawbone CBCT scans was acquired from patients who underwent implant placement with or without coronal restoration. The dataset was randomly divided into three subsets: training set (n = 225), validation set (n = 25) and testing set (n = 30). A CNN model was developed and trained using expert-based semi-automated segmentation (SS) of the implant and attached prosthetic crown as the ground truth. The performance of AS was assessed by comparing with SS and manually corrected automated segmentation referred to as refined-automated segmentation (R-AS). Evaluation metrics included timing, voxel-wise comparison based on confusion matrix and 3D surface differences. ResultsThe average time required for AS was 60 times faster (<30 s) than the SS approach. The CNN model was highly effective in segmenting dental implants both with and without coronal restoration, achieving a high dice similarity coefficient score of 0.92±0.02 and 0.91±0.03, respectively. Moreover, the root mean square deviation values were also found to be low (implant only: 0.08±0.09 mm, implant+restoration: 0.11±0.07 mm) when compared with R-AS, implying high AI segmentation accuracy. ConclusionsThe proposed cloud-based deep learning tool demonstrated high performance and time-efficient segmentation of implants on CBCT images. Clinical significanceAI-based segmentation of implants and prosthetic crowns can minimize the negative impact of artifacts and enhance the generalizability of creating dental virtual models. Furthermore, incorporating the suggested tool into existing CNN models specialized for segmenting anatomical structures can improve pre-surgical planning for implants and post-operative assessment of peri‑implant bone levels.

Bony structure enhanced synthetic CT generation using Dixon sequences for pelvis MR-only radiotherapy.

MRI-only radiotherapy planning (MROP) is beneficial to patients by avoiding MRI/CT registration errors, simplifying the radiation treatment simulation workflow and reducing exposure to ionizing radiation. MRI is the primary imaging modality for soft tissue delineation. Treatment planning CTs (i.e., CT simulation scan) are redundant if a synthetic CT (sCT) can be generated from the MRI to provide the patient positioning and electron density information. Unsupervised deep learning (DL) models like CycleGAN are widely used in MR-to-sCT conversion, when paired patient CT and MR image datasets are not available for model training. However, compared to supervised DL models, they cannot guarantee anatomic consistency, especially around bone. The purpose of this work was to improve the sCT accuracy generated from MRI around bone for MROP. To generate more reliable bony structures on sCT images, we proposed to add bony structure constraints in the unsupervised CycleGAN model's loss function and leverage Dixon constructed fat and in-phase (IP) MR images. Dixon images provide better bone contrast than T2-weighted images as inputs to a modified multi-channel CycleGAN. A private dataset with a total of 31 prostate cancer patients were used for training (20) and testing (11). We compared model performance with and without bony structure constraints using single- and multi-channel inputs. Among all the models, multi-channel CycleGAN with bony structure constraints had the lowest mean absolute error, both inside the bone and whole body (50.7 and 145.2 HU). This approach also resulted in the highest Dice similarity coefficient (0.88) of all bony structures compared with the planning CT. Modified multi-channel CycleGAN with bony structure constraints, taking Dixon-constructed fat and IP images as inputs, can generate clinically suitable sCT images in both bone and soft tissue. The generated sCT images have the potential to be used for accurate dose calculation and patient positioning in MROP radiation therapy.

TPFR-Net: U-shaped model for lung nodule segmentation based on transformer pooling and dual-attention feature reorganization.

Accurate segmentation of lung nodules is the key to diagnosing the lesion type of lung nodule. The complex boundaries of lung nodules and the visual similarity to surrounding tissues make precise segmentation of lung nodules challenging. Traditional CNN based lung nodule segmentation models focus on extracting local features from neighboring pixels and ignore global contextual information, which is prone to incomplete segmentation of lung nodule boundaries. In the U-shaped encoder-decoder structure, variations of image resolution caused by up-sampling and down-sampling result in the loss of feature information, which reduces the reliability of output features. This paper proposes transformer pooling module and dual-attention feature reorganization module to effectively improve the above two defects. Transformer pooling module innovatively fuses the self-attention layer and pooling layer in the transformer, which compensates for the limitation of convolution operation, reduces the loss of feature information in the pooling process, and decreases the computational complexity of the Transformer significantly. Dual-attention feature reorganization module innovatively employs the dual-attention mechanism of channel and spatial to improve the sub-pixel convolution, minimizing the loss of feature information during up-sampling. In addition, two convolutional modules are proposed in this paper, which together with transformer pooling module form an encoder that can adequately extract local features and global dependencies. We use the fusion loss function and deep supervision strategy in the decoder to train the model. The proposed model has been extensively experimented and evaluated on the LIDC-IDRI dataset, the highest Dice Similarity Coefficient is 91.84 and the highest sensitivity is 92.66, indicating the model's comprehensive capability has surpassed state-of-the-art UTNet. The model proposed in this paper has superior segmentation performance for lung nodules and can provide a more in-depth assessment of lung nodules' shape, size, and other characteristics, which is of important clinical significance and application value to assist physicians in the early diagnosis of lung nodules.

Development and clinical utility analysis of a prostate zonal segmentation model on T2-weighted imaging: a multicenter study

ObjectivesTo automatically segment prostate central gland (CG) and peripheral zone (PZ) on T2-weighted imaging using deep learning and assess the model’s clinical utility by comparing it with a radiologist annotation and analyzing relevant influencing factors, especially the prostate zonal volume.MethodsA 3D U-Net-based model was trained with 223 patients from one institution and tested using one internal testing group (n = 93) and two external testing datasets, including one public dataset (ETDpub, n = 141) and one private dataset from two centers (ETDpri, n = 59). The Dice similarity coefficients (DSCs), 95th Hausdorff distance (95HD), and average boundary distance (ABD) were calculated to evaluate the model’s performance and further compared with a junior radiologist’s performance in ETDpub. To investigate factors influencing the model performance, patients’ clinical characteristics, prostate morphology, and image parameters in ETDpri were collected and analyzed using beta regression.ResultsThe DSCs in the internal testing group, ETDpub, and ETDpri were 0.909, 0.889, and 0.869 for CG, and 0.844, 0.755, and 0.764 for PZ, respectively. The mean 95HD and ABD were less than 7.0 and 1.3 for both zones. The U-Net model outperformed the junior radiologist, having a higher DSC (0.769 vs. 0.706) and higher intraclass correlation coefficient for volume estimation in PZ (0.836 vs. 0.668). CG volume and Magnetic Resonance (MR) vendor were significant influencing factors for CG and PZ segmentation.ConclusionsThe 3D U-Net model showed good performance for CG and PZ auto-segmentation in all the testing groups and outperformed the junior radiologist for PZ segmentation. The model performance was susceptible to prostate morphology and MR scanner parameters.

Automatic Segmentation of Hyperreflective Foci in OCT Images Based on Lightweight DBR Network.

Hyperreflective foci (HF) reflects inflammatory responses for fundus diseases such as diabetic macular edema (DME), retina vein occlusion (RVO), and central serous chorioretinopathy (CSC). Shown as high contrast and reflectivity in optical coherence tomography (OCT) images, automatic segmentation of HF in OCT images is helpful for the prognosis of fundus diseases. Previous traditional methods were time-consuming and required high computing power. Hence, we proposed a lightweight network to segment HF (with a speed of 57ms per OCT image, at least 150ms faster than other methods). Our framework consists of two stages: an NLM filter and patch-based split to preprocess images and a lightweight DBR neural network to segment HF automatically. Experimental results from 3000 OCT images of 300 patients (100 DME,100 RVO, and 100 CSC) revealed that our method achieved HF segmentation successfully. The DBR network had the area under curves dice similarity coefficient (DSC) of 83.65%, 76.43%, and 82.20% in segmenting HF in DME, RVO, and CSC on the test cohort respectively. Our DBR network achieves at least 5% higher DSC than previous methods. HF in DME was more easily segmented compared with the other two types. In addition, our DBR network is universally applicable to clinical practice with the ability to segment HF in a wide range of fundus diseases.

Radiomics and Artificial Intelligence Study of Masseter Muscle Segmentation in Patients With Hemifacial Microsomia.

Hemifacial microsomia (HFM) is one of the most common congenital craniofacial condition often accompanied by masseter muscle involvement. U-Net neural convolution network for masseter segmentation is expected to achieve an efficient evaluation of masseter muscle. A database was established with 108 patients with HFM from June 2012 to June 2019 in our center. Demographic data, OMENS classification, and 1-mm layer thick 3-dimensional computed tomography were included. Two radiologists manually segmented masseter muscles in a consensus reading as the ground truth. A test set of 20 cases was duplicated into 2 groups: an experimental group with the intelligent algorithm and a control group with manual segmentation. The U-net follows the design of 3D RoI-Aware U-Net with overlapping window strategy and references to our previous study of masseter segmentation in a healthy population system. Sorensen dice-similarity coefficient (DSC) muscle volume, average surface distance, recall, and time were used to validate compared with the ground truth. The mean DSC value of 0.794±0.028 for the experiment group was compared with the manual segmentation (0.885±0.118) with α=0.05 and a noninferiority margin of 15%. In addition, higher DSC was reported in patients with milder mandible deformity ( r =0.824, P <0.05). Moreover, intelligent automatic segmentation takes only 6.4 seconds showing great efficiency. We first proposed a U-net neural convolutional network and achieved automatic segmentation of masseter muscles in patients with HFM. It is a great attempt at intelligent diagnosis and evaluation of craniofacial diseases.

CTBP-Net: Lung nodule segmentation model based on the cross-transformer and bidirectional pyramid

Accurate segmentation of lung nodules is crucial for subsequent pathological analysis and diagnosis. However, precise segmentation is difficult due to their complex morphological features and visual characteristics similar to surrounding tissues. The lung nodule segmentation model based on a convolutional neural network focuses on extracting local features but ignores global features. Meanwhile, the shallow layer of the encoder-decoder structure is rich in spatial information but lacks semantic information, and the deep layer has rich semantic information but lacks spatial information. This paper innovatively proposes a Cross-Transformer module and a Bidirectional Pyramid module to effectively ameliorate the above two defects. The former combines the Transformer's internal and external self-attention to alleviate the limitations of convolution operations and enhance the ability of the Transformer to extract global features. The latter adds bidirectional channels to the feature pyramid to increase the interaction of feature information between the shallow and deep layers of the model, reducing the differences in features at different stages of the model. Extensive experiments of lung nodule segmentation performed on the LIDC-IDRI dataset have shown that the highest Dice Similarity Coefficient is 91.68, the highest Sensitivity is 92.24, and the comprehensive performance has surpassed UTNet. The model proposed in this paper has superior performance for the segmentation of lung nodules and can further analyze the shape, size, and other characteristics of lung nodules, which is of great clinical significance and application value to physicians in the early diagnosis of lung nodules.