Fully Automatic Deep Learning Model for Spine Refracture in Patients with OVCF: A Multi-Center Study.

BACKGROUND: 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). OBJECTIVE: 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. METHODS: 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. RESULTS: 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. CONCLUSION: 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.

Training and assessing convolutional neural network performance in automatic vascular segmentation using Ga-68 DOTATATE PET/CT.

To evaluate a convolutional neural network's performance (nnU-Net) in the assessment of vascular contours, calcification and PET tracer activity using Ga-68 DOTATATE PET/CT. Patients who underwent Ga-68 DOTATATE PET/CT imaging over a 12-month period for neuroendocrine investigation were included. Manual cardiac and aortic segmentations were performed by an experienced observer. Scans were randomly allocated in ratio 64:16:20 for training, validation and testing of the nnU-Net model. PET tracer uptake and calcium scoring were compared between segmentation methods and different observers. 116 patients (53.5% female) with a median age of 64.5 years (range 23-79) were included. There were strong, positive correlations between all segmentations (mostly r > 0.98). There were no significant differences between manual and AI segmentation of SUVmean for global cardiac (mean ± SD 0.71 ± 0.22 vs. 0.71 ± 0.22; mean diff 0.001 ± 0.008, p > 0.05), ascending aorta (mean ± SD 0.44 ± 0.14 vs. 0.44 ± 0.14; mean diff 0.002 ± 0.01, p > 0.05), aortic arch (mean ± SD 0.44 ± 0.10 vs. 0.43 ± 0.10; mean diff 0.008 ± 0.16, p > 0.05) and descending aorta (mean ± SD < 0.001; 0.58 ± 0.12 vs. 0.57 ± 0.12; mean diff 0.01 ± 0.03, p > 0.05) contours. There was excellent agreement between the majority of manual and AI segmentation measures (r ≥ 0.80) and in all vascular contour calcium scores. Compared with the manual segmentation approach, the CNN required a significantly lower workflow time. AI segmentation of vascular contours using nnU-Net resulted in very similar measures of PET tracer uptake and vascular calcification when compared to an experienced observer and significantly reduced workflow time.

Automated segmentation of brain metastases with deep learning: a multi-center, randomized crossover, multi-reader evaluation study.

BACKGROUND: Artificial intelligence has been proposed for brain metastasis (BM) segmentation but it has not been fully clinically validated. The aim of this study was to develop and evaluate a system for BM segmentation. METHODS: A deep-learning-based BM segmentation system (BMSS) was developed using contrast-enhanced MR images from 488 patients with 10,338 brain metastases. A randomized crossover, multi-reader study was then conducted to evaluate the performance of the BMSS for BM segmentation using data prospectively collected from 50 patients with 203 metastases at five centers. Five radiology residents and five attending radiologists were randomly assigned to contour the same prospective set in assisted and unassisted modes. Aided and unaided Dice similarity coefficients (DSCs) and contouring times per lesion were compared. RESULTS: The BMSS alone yielded a median DSC of 0.91 (95% confidence interval, 0.90-0.92) in the multi-center set and showed comparable performance between the internal and external sets (p = 0.67). With BMSS assistance, the readers increased the median DSC from 0.87 (0.87-0.88) to 0.92 (0.92-0.92) (p < 0.001) with a median time saving of 42% (40-45%) per lesion. Resident readers showed a greater improvement than attending readers in contouring accuracy (improved median DSC, 0.05 [0.05-0.05] vs. 0.03 [0.03-0.03]; p < 0.001), but a similar time reduction (reduced median time, 44% [40-47%] vs. 40% [37-44%]; p = 0.92) with BMSS assistance. CONCLUSIONS: The BMSS can be optimally applied to improve the efficiency of brain metastasis delineation in clinical practice.

Advanced OCTA imaging segmentation: Unsupervised, non-linear retinal vessel detection using modified self-organizing maps and joint MGRF modeling.

BACKGROUND AND OBJECTIVE: This paper proposes a fully automated and unsupervised stochastic segmentation approach using two-level joint Markov-Gibbs Random Field (MGRF) to detect the vascular system from retinal Optical Coherence Tomography Angiography (OCTA) images, which is a critical step in developing Computer-Aided Diagnosis (CAD) systems for detecting retinal diseases. METHODS: Using a new probabilistic model based on a Linear Combination of Discrete Gaussian (LCDG), the first level models the appearance of OCTA images and their spatially smoothed images. The parameters of the LCDG model are estimated using a modified Expectation Maximization (EM) algorithm. The second level models the maps of OCTA images, including the vascular system and other retina tissues, using MGRF with analytically estimated parameters from the input images. The proposed segmentation approach employs modified self-organizing maps as a MAP-based optimizer maximizing the joint likelihood and handles the Joint MGRF model in a new, unsupervised way. This approach deviates from traditional stochastic optimization approaches and leverages non-linear optimization to achieve more accurate segmentation results. RESULTS: The proposed segmentation framework is evaluated quantitatively on a dataset of 204 subjects. Achieving 0.92 ± 0.03 Dice similarity coefficient, 0.69 ± 0.25 95-percentile bidirectional Hausdorff distance, and 0.93 ± 0.03 accuracy, confirms the superior performance of the proposed approach. CONCLUSIONS: The conclusions drawn from the study highlight the superior performance of the proposed unsupervised and fully automated segmentation approach in detecting the vascular system from OCTA images. This approach not only deviates from traditional methods but also achieves more accurate segmentation results, demonstrating its potential in aiding the development of CAD systems for detecting retinal diseases.

Enhanced choroid plexus segmentation with 3D UX-Net and its association with disease progression in multiple sclerosis.

BACKGROUND: The choroid plexus (CP) is suggested to be closely associated with the neuroinflammation of multiple sclerosis (MS). Segmentation based on deep learning (DL) could facilitate rapid and reproducible volume assessment of the CP, which is crucial for elucidating its role in MS. PURPOSE: To develop a reliable DL model for the automatic segmentation of CP, and further validate its clinical significance in MS. METHODS: The 3D UX-Net model (3D U-Net used for comparison) was trained and validated on T1-weighted MRI from a cohort of 216 relapsing-remitting MS (RRMS) patients and 75 healthy subjects. Among these, 53 RRMS with baseline and 2-year follow-up scans formed an internal test set (dataset1b). Another 58 RRMS from multi-center data served as an external test set (dataset2). Dice coefficient was computed to assess segmentation performance. Compare the correlation of CP volume obtained through automatic and manual segmentation with clinical outcomes in MS. Disability and cognitive function of patients were assessed using the Expanded Disability Status Scale (EDSS) and Symbol Digit Modalities Test (SDMT). RESULTS: The 3D UX-Net model achieved Dice coefficients of 0.875 ± 0.030 and 0.870 ± 0.044 for CP segmentation on dataset1b and dataset2, respectively, outperforming 3D U-Net's scores of 0.809 ± 0.098 and 0.601 ± 0.226. Furthermore, CP volumes segmented by the 3D UX-Net model aligned consistently with clinical outcomes compared to manual segmentation. In dataset1b, both manual and automatic segmentation revealed a significant positive correlation between normalized CP volume (nCPV) and EDSS scores at baseline (manual: r = 0.285, p = 0.045; automatic: r = 0.287, p = 0.044) and a negative correlation with SDMT scores (manual: r = -0.331, p = 0.020; automatic: r = -0.329, p = 0.021). In dataset2, similar correlations were found with EDSS scores (manual: r = 0.337, p = 0.021; automatic: r = 0.346, p = 0.017). Meanwhile, in dataset1b, both manual and automatic segmentation revealed a significant increase in nCPV from baseline to follow-up (p < 0.05). The increase of nCPV was more pronounced in patients with disability worsened than stable patients (manual: p = 0.023; automatic: p = 0.018). Patients receiving disease-modifying therapy (DMT) exhibited a significantly lower nCPV increase than untreated patients (manual: p = 0.004; automatic: p = 0.004). CONCLUSION: The 3D UX-Net model demonstrated strong segmentation performance for the CP, and the automatic segmented CP can be directly used in MS clinical practice. CP volume can serve as a surrogate imaging biomarker for monitoring disease progression and DMT response in MS patients.

Charged Gold Nanoparticles for Target Identification-Alignment and Automatic Segmentation of CT Image-Guided Adaptive Radiotherapy in Small Hepatocellular Carcinoma.

Because of the challenges posed by anatomical uncertainties and the low resolution of plain computed tomography (CT) scans, implementing adaptive radiotherapy (ART) for small hepatocellular carcinoma (sHCC) using artificial intelligence (AI) faces obstacles in tumor identification-alignment and automatic segmentation. The current study aims to improve sHCC imaging for ART using a gold nanoparticle (Au NP)-based CT contrast agent to enhance AI-driven automated image processing. The synthesized charged Au NPs demonstrated notable in vitro aggregation, low cytotoxicity, and minimal organ toxicity. Over time, an in situ sHCC mouse model was established for in vivo CT imaging at multiple time points. The enhanced CT images processed using 3D U-Net and 3D Trans U-Net AI models demonstrated high geometric and dosimetric accuracy. Therefore, charged Au NPs enable accurate and automatic sHCC segmentation in CT images using classical AI models, potentially addressing the technical challenges related to tumor identification, alignment, and automatic segmentation in CT-guided online ART.

Improved automatic segmentation of brain metastasis gross tumor volume in computed tomography images for radiotherapy: a position attention module for U-Net architecture.

Background: Brain metastases present significant challenges in radiotherapy due to the need for precise tumor delineation. Traditional methods often lack the efficiency and accuracy required for optimal treatment planning. This paper proposes an improved U-Net model that uses a position attention module (PAM) for automated segmentation of gross tumor volumes (GTVs) in computed tomography (CT) simulation images of patients with brain metastases to improve the efficiency and accuracy of radiotherapy planning and segmentation. Methods: We retrospectively collected CT simulation imaging datasets of patients with brain metastases from two centers, which were designated as the training and external validation datasets. The U-Net architecture was enhanced by incorporating a PAM into the transition layer, which improved the automated segmentation capability of the U-Net model. With cross-entropy loss employed as the loss function, the samples from the training dataset underwent training. The model's segmentation performance on the external validation dataset was assessed using metrics including the Dice similarity coefficient (DSC), intersection over union (IoU), accuracy, sensitivity, specificity, Matthews correlation coefficient (MCC), and Hausdorff distance (HD). Results: The proposed automated segmentation model demonstrated promising performance on the external validation dataset, achieving a DSC of 0.753±0.172. In terms of evaluation metrics (including the DSC, IoU, accuracy, sensitivity, MCC, and HD), the model outperformed the standard U-Net, which had a DSC of 0.691±0.142. The proposed model produced segmentation results that were closer to the ground truth and could reveal more detailed features of brain metastases. Conclusions: The PAM-improved U-Net model offers considerable advantages in the automated segmentation of the GTV in CT simulation images for patients with brain metastases. Its superior performance in comparison with the standard U-Net model supports its potential for streamlining and improving the accuracy of radiotherapy. With its ability to produce segmentation results consistent with the ground truth, the proposed model holds promise for clinical adoption and provides a reference for radiation oncologists to make more informed GTV segmentation decisions.

The Quantitative Evaluation of Automatic Segmentation in Lumbar Magnetic Resonance Images.

OBJECTIVE: This study aims to overcome challenges in lumbar spine imaging, particularly lumbar spinal stenosis, by developing an automated segmentation model using advanced techniques. Traditional manual measurement and lesion detection methods are limited by subjectivity and inefficiency. The objective is to create an accurate and automated segmentation model that identifies anatomical structures in lumbar spine magnetic resonance imaging scans. METHODS: Leveraging a dataset of 539 lumbar spinal stenosis patients, the study utilizes the residual U-Net for semantic segmentation in sagittal and axial lumbar spine magnetic resonance images. The model, trained to recognize specific tissue categories, employs a geometry algorithm for anatomical structure quantification. Validation metrics, like Intersection over Union (IOU) and Dice coefficients, validate the residual U-Net's segmentation accuracy. A novel rotation matrix approach is introduced for detecting bulging discs, assessing dural sac compression, and measuring yellow ligament thickness. RESULTS: The residual U-Net achieves high precision in segmenting lumbar spine structures, with mean IOU values ranging from 0.82 to 0.93 across various tissue categories and views. The automated quantification system provides measurements for intervertebral disc dimensions, dural sac diameter, yellow ligament thickness, and disc hydration. Consistency between training and testing datasets assures the robustness of automated measurements. CONCLUSION: Automated lumbar spine segmentation with residual U-Net and deep learning exhibits high precision in identifying anatomical structures, facilitating efficient quantification in lumbar spinal stenosis cases. The introduction of a rotation matrix enhances lesion detection, promising improved diagnostic accuracy, and supporting treatment decisions for lumbar spinal stenosis patients.

Feasibility Study on Constructing Dosimetric Correlated Geometric Parameters for Automatic Segmentation Evaluation.

PURPOSE: To investigate the feasibility of constructing new geometric parameters that correlate well with dosimetric parameters. METHODS: 100 rectal cancer patients were enrolled. The targets were identified manually, while the organs at risk (bladder, small bowel, left and right femoral heads) were segmented both manually and automatically. The radiotherapy plans were optimized according to the automatically contoured organs at risk. Forty cases were randomly selected to establish the relationship between dose and distance for each organ at risk, termed "dose-distance curves," which were then applied to the new geometric parameters. The correlation between these new geometric parameters and dosimetric parameters was analyzed in the remaining 60 test cases. RESULTS: The "dose-distance curves" were similar across the four organs at risk, exhibiting an inverse function shape with a rapid decrease initially and a slower rate at a later stage. The Pearson correlation coefficients of new geometric parameters and dosimetric parameters in the bladder, small intestine, and left and right femur heads were 0.96, 0.97, 0.88, and 0.70, respectively. CONCLUSIONS: The new geometric parameters predicated on "distance from the target" showed a high correlation with corresponding dosimetric parameters in rectal cancer cases. It is feasible to utilize the new geometric parameters to evaluate the dose deviation attributable to automatic segmentation.

Rapid segmentation of computed tomography angiography images of the aortic valve: the efficacy and clinical value of a deep learning algorithm.

Objectives: The goal of this study was to explore the reliability and clinical value of fast, accurate automatic segmentation of the aortic root based on a deep learning tool compared with computed tomography angiography. Methods: A deep learning tool for automatic 3-dimensional aortic root reconstruction, the CVPILOT system (TAVIMercy Data Technology Ltd., Nanjing, China), was trained and tested using computed tomography angiography scans collected from 183 patients undergoing transcatheter aortic valve replacement from January 2021 to December 2022. The quality of the reconstructed models was assessed using validation data sets and evaluated clinically by experts. Results: The segmentation of the ascending aorta and the left ventricle attained Dice similarity coefficients (DSC) of 0.9806/0.9711 and 0.9603/0.9643 for the training and validation sets, respectively. The leaflets had a DSC of 0.8049/0.7931, and the calcification had a DSC of 0.8814/0.8630. After 6 months of application, the system modeling time was reduced to 19.83 s. Conclusion: For patients undergoing transcatheter aortic valve replacement, the CVPILOT system facilitates clinical workflow. The reliable evaluation quality of the platform indicates broad clinical application prospects in the future.

High-Resolution Magnetic Resonance Neurography at 7T: A Pilot Study of Hand Innervation.

The emergence of 7T clinical MRI technology has sparked our interest in its ability to discern the complex structures of the hand. Our primary objective was to assess the sensory and motor nerve structures of the hand, specifically nerves and Pacinian corpuscles, with the dual purpose of aiding diagnostic endeavors and supporting reconstructive surgical procedures. Ethical approval was obtained to carry out 7T MRI scans on a cohort of volunteers. Four volunteers assumed a prone position, with their hands (N = 8) positioned in a "superman" posture. To immobilize and maintain the hand in a strictly horizontal position, it was affixed to a plastic plate. Passive B0 shimming was implemented. Once high-resolution 3D images had been acquired using a multi-transmit head coil, advanced post-processing techniques were used to meticulously delineate the nerve fiber networks and mechanoreceptors. Across all participants, digital nerves were consistently located on the phalanges area, on average, between 2.5 and 3.5 mm beneath the skin, except within flexion folds where the nerve was approximately 1.8 mm from the surface. On the phalanges area, the mean distance from digital nerves to joints was approximately 1.5 mm. The nerves of the fingers were closer to the bone than to the surface of the skin. Furthermore, Pacinian corpuscles exhibited a notable clustering primarily within the metacarpal zone, situated on the palmar aspect. Our study yielded promising results, successfully reconstructing and meticulously describing the anatomy of nerve fibers spanning from the carpus to the digital nerve division, alongside the identification of Pacinian corpuscles, in four healthy volunteers (eight hands).

Automatic segmentation of MRI in prospective breast volume evaluation: Comparison of different assessments for immediate breast reconstruction.

BACKGROUND: Assessment of breast volume is essential in preoperative planning of immediate breast reconstruction (IBR) surgery to achieve satisfactory cosmetic outcome. This study introduced a breast volume measurement tool that can be used to perform automatic segmentation of magnetic resonance images (MRI) and calculation of breast volume. We compared the accuracy and reliability of this measurement method with four other conventional modalities. METHODS: Patients who were scheduled to undergo mastectomy with IBR between 2016 and 2021 were enrolled in the study. Five different breast volume assessments, including automatic segmentation of MRI, manual segmentation of MRI, 3D surface imaging, mammography, and the BREAST-V formula, were used to evaluate different breast volumes. The results were validated using water displacement volumes of the mastectomy specimens. RESULTS: In this pilot study, a total of 50 female patients met the inclusion criteria and contributed 54 breast specimens to the volumetric analysis. There was a strong linear association between the MRI and water displacement methods (automatic segmentation: r = 0.911, p < 0.001; manual segmentation: r = 0.924, p < 0.001), followed by 3D surface imaging (r = 0.858, p < 0.001), mammography (r = 0.841, p < 0.001), and Breast-V formula (r = 0.838, p < 0.001). Breast volumes measured using automatic and manual segmentation of MRI had lower mean relative errors (30.3% ± 22.0% and 28.9% ± 19.8, respectively) than 3D surface imaging (38.9% ± 31.2), Breast-V formula (44.8% ± 25.8), and mammography (60.3% ± 37.6). CONCLUSION: Breast volume assessment using the MRI methods had better accuracy and reliability than the other methods used in our study. Breast volume measurement using automatic segmentation of MRI could be more efficient compared to the conventional methods.

Accuracy of artificial intelligence-based segmentation of the mandibular canal in CBCT.

OBJECTIVES: To investigate the accuracy of artificial intelligence (AI)-based segmentation of the mandibular canal, compared to the conventional manual tracing, implementing implant planning software. MATERIALS AND METHODS: Localization of the mandibular canals was performed for 104 randomly selected patients. A localization was performed by three experienced clinicians in order to serve as control. Five tracings were performed: One from a clinician with a moderate experience with a manual tracing (I1), followed by the implementation of an automatic refinement (I2), one manual from a dental student (S1), and one from the experienced clinician, followed by an automatic refinement (E). Subsequently, two fully automatic AI-driven segmentations were performed (A1,A2). The accuracy between each method was measured using root mean square error calculation. RESULTS: The discrepancy among the models of the mandibular canals, between the experienced clinicians and each investigated method ranged from 0.21 to 7.65 mm with a mean of 3.5 mm RMS error. The analysis of each separate mandibular canal's section revealed that mean RMS error was higher in the posterior and anterior loop compared to the middle section. Regarding time efficiency, tracing by experienced users required more time compared to AI-driven segmentation. CONCLUSIONS: The experience of the clinician had a significant influence on the accuracy of mandibular canal's localization. An AI-driven segmentation of the mandibular canal constitutes a time-efficient and reliable procedure for pre-operative implant planning. Nevertheless, AI-based segmentation results should always be verified, as a subsequent manual refinement of the initial segmentation may be required to avoid clinical significant errors.

Deep learning automatic semantic segmentation of glioblastoma multiforme regions on multimodal magnetic resonance images.

OBJECTIVES: In patients having naïve glioblastoma multiforme (GBM), this study aims to assess the efficacy of Deep Learning algorithms in automating the segmentation of brain magnetic resonance (MR) images to accurately determine 3D masks for 4 distinct regions: enhanced tumor, peritumoral edema, non-enhanced/necrotic tumor, and total tumor. MATERIAL AND METHODS: A 3D U-Net neural network algorithm was developed for semantic segmentation of GBM. The training dataset was manually delineated by a group of expert neuroradiologists on MR images from the Brain Tumor Segmentation Challenge 2021 (BraTS2021) image repository, as ground truth labels for diverse glioma (GBM and low-grade glioma) subregions across four MR sequences (T1w, T1w-contrast enhanced, T2w, and FLAIR) in 1251 patients. The in-house test was performed on 50 GBM patients from our cohort (PerProGlio project). By exploring various hyperparameters, the network's performance was optimized, and the most optimal parameter configuration was identified. The assessment of the optimized network's performance utilized Dice scores, precision, and sensitivity metrics. RESULTS: Our adaptation of the 3D U-net with additional residual blocks demonstrated reliable performance on both the BraTS2021 dataset and the in-house PerProGlio cohort, employing only T1w-ce sequences for enhancement and non-enhanced/necrotic tumor models and T1w-ce + T2w + FLAIR for peritumoral edema and total tumor. The mean Dice scores (training and test) were 0.89 and 0.75; 0.75 and 0.64; 0.79 and 0.71; and 0.60 and 0.55, for total tumor, edema, enhanced tumor, and non-enhanced/necrotic tumor, respectively. CONCLUSIONS: The results underscore the high precision with which our network can effectively segment GBM tumors and their distinct subregions. The level of accuracy achieved agrees with the coefficients recorded in previous GBM studies. In particular, our approach allows model specialization for each of the different tumor subregions employing only those MR sequences that provide value for segmentation.

Automatic gross tumor volume segmentation with failure detection for safe implementation in locally advanced cervical cancer.

Background and Purpose: Automatic segmentation methods have greatly changed the RadioTherapy (RT) workflow, but still need to be extended to target volumes. In this paper, Deep Learning (DL) models were compared for Gross Tumor Volume (GTV) segmentation in locally advanced cervical cancer, and a novel investigation into failure detection was introduced by utilizing radiomic features. Methods and materials: We trained eight DL models (UNet, VNet, SegResNet, SegResNetVAE) for 2D and 3D segmentation. Ensembling individually trained models during cross-validation generated the final segmentation. To detect failures, binary classifiers were trained using radiomic features extracted from segmented GTVs as inputs, aiming to classify contours based on whether their Dice Similarity Coefficient ( DSC ) < T and DSC ⩾ T . Two distinct cohorts of T2-Weighted (T2W) pre-RT MR images captured in 2D sequences were used: one retrospective cohort consisting of 115 LACC patients from 30 scanners, and the other prospective cohort, comprising 51 patients from 7 scanners, used for testing. Results: Segmentation by 2D-SegResNet achieved the best DSC, Surface DSC ( SDSC 3 mm ), and 95th Hausdorff Distance (95HD): DSC = 0.72 ± 0.16, SDSC 3 mm =0.66 ± 0.17, and 95HD = 14.6 ± 9.0 mm without missing segmentation ( M =0) on the test cohort. Failure detection could generate precision ( P = 0.88 ), recall ( R = 0.75 ), F1-score ( F = 0.81 ), and accuracy ( A = 0.86 ) using Logistic Regression (LR) classifier on the test cohort with a threshold T = 0.67 on DSC values. Conclusions: Our study revealed that segmentation accuracy varies slightly among different DL methods, with 2D networks outperforming 3D networks in 2D MRI sequences. Doctors found the time-saving aspect advantageous. The proposed failure detection could guide doctors in sensitive cases.

Artificial intelligence in radiotherapy: Current applications and future trends.

Radiation therapy has dramatically changed with the advent of computed tomography and intensity modulation. This added complexity to the workflow but allowed for more precise and reproducible treatment. As a result, these advances required the accurate delineation of many more volumes, raising questions about how to delineate them, in a uniform manner across centers. Then, as computing power improved, reverse planning became possible and three-dimensional dose distributions could be generated. Artificial intelligence offers the opportunity to make such workflow more efficient while increasing practice homogeneity. Many artificial intelligence-based tools are being implemented in routine practice to increase efficiency, reduce workload and improve homogeneity of treatments. Data retrieved from this workflow could be combined with clinical data and omic data to develop predictive tools to support clinical decision-making process. Such predictive tools are at the stage of proof-of-concept and need to be explainatory, prospectively validated, and based on large and multicenter cohorts. Nevertheless, they could bridge the gap to personalized radiation oncology, by personalizing oncologic strategies, dose prescriptions to tumor volumes and dose constraints to organs at risk.

Use of Artificial Intelligence With Deep Learning Approaches for the Follow-up of Infrarenal Endovascular Aortic Repair.

INTRODUCTION: Endoleaks represent one of the main complications after endovascular aortic repair (EVAR) and can lead to increased re-intervention rates and secondary rupture. Serial lifelong surveillance is required and traditionally involves cross-sectional imaging with manual axial measurements. Artificial intelligence (AI)-based imaging analysis has been developed and may provide a more precise and faster assessment. This study aims to evaluate the ability of an AI-based software to assess post-EVAR morphological changes over time, detect endoleaks, and associate them with EVAR-related adverse events. METHODS: Patients who underwent EVAR at a tertiary hospital from January 2017 to March 2020 with at least 2 follow-up computed tomography angiography (CTA) were analyzed using PRAEVAorta 2 (Nurea). The software was compared to the ground truth provided by human experts using Sensitivity (Se), Specificity (Sp), Negative Predictive Value (NPV), and Positive Predictive Value (PPV). Endovascular aortic repair-related adverse events were defined as aneurysm-related death, rupture, endoleak, limb occlusion, and EVAR-related re-interventions. RESULTS: Fifty-six patients were included with a median imaging follow-up of 27 months (interquartile range [IQR]: 20-40). There were no significant differences overtime in the evolution of maximum aneurysm diameters (55.62 mm [IQR: 52.33-59.25] vs 54.34 mm [IQR: 46.13-59.47]; p=0.2162) or volumes (130.4 cm3 [IQR: 113.8-171.7] vs 125.4 cm3 [IQR: 96.3-169.1]; p=0.1131) despite a -13.47% decrease in the volume of thrombus (p=0.0216). PRAEVAorta achieved a Se of 89.47% (95% confidence interval [CI]: 80.58 to 94.57), a Sp of 91.25% (95% CI: 83.02 to 95.70), a PPV of 90.67% (95% CI: 81.97 to 95.41), and an NPV of 90.12% (95% CI: 81.70 to 94.91) in detecting endoleaks. Endovascular aortic repair-related adverse events were associated with global volume modifications with an area under the curve (AUC) of 0.7806 vs 0.7277 for maximum diameter. The same trend was observed for endoleaks (AUC of 0.7086 vs 0.6711). CONCLUSIONS: The AI-based software PRAEVAorta enabled a detailed anatomic characterization of aortic remodeling post-EVAR and showed its potential interest for automatic detection of endoleaks during follow-up. The association of aortic aneurysmal volume with EVAR-related adverse events and endoleaks was more robust compared with maximum diameter. CLINICAL IMPACT: The integration of PRAEVAorta AI software into clinical practice promises a transformative shift in post-EVAR surveillance. By offering precise and rapid detection of endoleaks and comprehensive anatomic assessments, clinicians can expect enhanced diagnostic accuracy and streamlined patient management. This innovation reduces reliance on manual measurements, potentially reducing interpretation errors and shortening evaluation times. Ultimately, PRAEVAorta's capabilities hold the potential to optimize patient care, leading to more timely interventions and improved outcomes in endovascular aortic repair.

Deep learning based automatic segmentation of the Internal Pudendal Artery in definitive radiotherapy treatment planning of localized prostate cancer.

Background and purpose: Radiation-induced erectile dysfunction (RiED) commonly affects prostate cancer patients, prompting clinical trials across institutions to explore dose-sparing to internal-pudendal-arteries (IPA) for preserving sexual potency. IPA, challenging to segment, isn't conventionally considered an organ-at-risk (OAR). This study proposes a deep learning (DL) auto-segmentation model for IPA, using Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) or CT alone to accommodate varied clinical practices. Materials and methods: A total of 86 patients with CT and MRI images and noisy IPA labels were recruited in this study. We split the data into 42/14/30 for model training, testing, and a clinical observer study, respectively. There were three major innovations in this model: 1) we designed an architecture with squeeze-and-excite blocks and modality attention for effective feature extraction and production of accurate segmentation, 2) a novel loss function was used for training the model effectively with noisy labels, and 3) modality dropout strategy was used for making the model capable of segmentation in the absence of MRI. Results: Test dataset metrics were DSC 61.71 ± 7.7 %, ASD 2.5 ± .87 mm, and HD95 7.0 ± 2.3 mm. AI segmented contours showed dosimetric similarity to expert physician's contours. Observer study indicated higher scores for AI contours (mean = 3.7) compared to inexperienced physicians' contours (mean = 3.1). Inexperienced physicians improved scores to 3.7 when starting with AI contours. Conclusion: The proposed model achieved good quality IPA contours to improve uniformity of segmentation and to facilitate introduction of standardized IPA segmentation into clinical trials and practice.

Automatic left ventricle volume and mass quantification from 2D cine-MRI: Investigating papillary muscle influence.

OBJECTIVE: Early detection of cardiovascular diseases is based on accurate quantification of the left ventricle (LV) function parameters. In this paper, we propose a fully automatic framework for LV volume and mass quantification from 2D-cine MR images already segmented using U-Net. METHODS: The general framework consists of three main steps: Data preparation including automatic LV localization using a convolution neural network (CNN) and application of morphological operations to exclude papillary muscles from the LV cavity. The second step consists in automatically extracting the LV contours using U-Net architecture. Finally, by integrating temporal information which is manifested by a spatial motion of myocytes as a third dimension, we calculated LV volume, LV ejection fraction (LVEF) and left ventricle mass (LVM). Based on these parameters, we detected and quantified cardiac contraction abnormalities using Python software. RESULTS: CNN was trained with 35 patients and tested on 15 patients from the ACDC database with an accuracy of 99,15 %. U-Net architecture was trained using ACDC database and evaluated using local dataset with a Dice similarity coefficient (DSC) of 99,78 % and a Hausdorff Distance (HD) of 4.468 mm (p < 0,001). Quantification results showed a strong correlation with physiological measures with a Pearson correlation coefficient (PCC) of 0,991 for LV volume, 0.962 for LVEF, 0.98 for stroke volume (SV) and 0.923 for LVM after pillars' elimination. Clinically, our method allows regional and accurate identification of pathological myocardial segments and can serve as a diagnostic aid tool of cardiac contraction abnormalities. CONCLUSION: Experimental results prove the usefulness of the proposed method for LV volume and function quantification and verify its potential clinical applicability.

Improved brain metastases segmentation using generative adversarial network and conditional random field optimization mask R-CNN.

BACKGROUND: In radiotherapy, the delineation of the gross tumor volume (GTV) in brain metastases using computed tomography (CT) simulation localization is very important. However, despite the criticality of this process, a pronounced gap exists in the availability of tools tailored for the automatic segmentation of the GTV based on CT simulation localization images. PURPOSE: This study aims to fill this gap by devising an effective tool specifically for the automatic segmentation of the GTV using CT simulation localization images. METHODS: A dual-network generative adversarial network (GAN) architecture was developed, wherein the generator focused on refining CT images for more precise delineation, and the discriminator differentiated between real and augmented images. This architecture was coupled with the Mask R-CNN model to achieve meticulous GTV segmentation. An end-to-end training process facilitated the integration between the GAN and Mask R-CNN functionalities. Furthermore, a conditional random field (CRF) was incorporated to refine the initial masks generated by the Mask R-CNN model to ensure optimal segmentation accuracy. The performance was assessed using key metrics, namely, the Dice coefficient (DSC), intersection over union (IoU), accuracy, specificity, and sensitivity. RESULTS: The GAN+Mask R-CNN+CRF integration method in this study performs well in GTV segmentation. In particular, the model has an overall average DSC of 0.819 ± 0.102 and an IoU of 0.712 ± 0.111 in the internal validation. The overall average DSC in the external validation data is 0.726 ± 0.128 and the IoU is 0.640 ± 0.136. It demonstrates favorable generalization ability. CONCLUSION: The integration of the GAN, Mask R-CNN, and CRF optimization provides a pioneering tool for the sophisticated segmentation of the GTV in brain metastases using CT simulation localization images. The method proposed in this study can provide a robust automatic segmentation approach for brain metastases in the absence of MRI.