Deformable registration of preoperative MR and intraoperative long-length tomosynthesis images for guidance of spine surgery via image synthesis.

PURPOSE: Improved integration and use of preoperative imaging during surgery hold significant potential for enhancing treatment planning and instrument guidance through surgical navigation. Despite its prevalent use in diagnostic settings, MR imaging is rarely used for navigation in spine surgery. This study aims to leverage MR imaging for intraoperative visualization of spine anatomy, particularly in cases where CT imaging is unavailable or when minimizing radiation exposure is essential, such as in pediatric surgery. METHODS: This work presents a method for deformable 3D-2D registration of preoperative MR images with a novel intraoperative long-length tomosynthesis imaging modality (viz., Long-Film [LF]). A conditional generative adversarial network is used to translate MR images to an intermediate bone image suitable for registration, followed by a model-based 3D-2D registration algorithm to deformably map the synthesized images to LF images. The algorithm's performance was evaluated on cadaveric specimens with implanted markers and controlled deformation, and in clinical images of patients undergoing spine surgery as part of a large-scale clinical study on LF imaging. RESULTS: The proposed method yielded a median 2D projection distance error of 2.0 mm (interquartile range [IQR]: 1.1-3.3 mm) and a 3D target registration error of 1.5 mm (IQR: 0.8-2.1 mm) in cadaver studies. Notably, the multi-scale approach exhibited significantly higher accuracy compared to rigid solutions and effectively managed the challenges posed by piecewise rigid spine deformation. The robustness and consistency of the method were evaluated on clinical images, yielding no outliers on vertebrae without surgical instrumentation and 3% outliers on vertebrae with instrumentation. CONCLUSIONS: This work constitutes the first reported approach for deformable MR to LF registration based on deep image synthesis. The proposed framework provides access to the preoperative annotations and planning information during surgery and enables surgical navigation within the context of MR images and/or dual-plane LF images.

Active Learning in Brain Tumor Segmentation with Uncertainty Sampling and Annotation Redundancy Restriction.

Deep learning models have demonstrated great potential in medical imaging but are limited by the expensive, large volume of annotations required. To address this, we compared different active learning strategies by training models on subsets of the most informative images using real-world clinical datasets for brain tumor segmentation and proposing a framework that minimizes the data needed while maintaining performance. Then, 638 multi-institutional brain tumor magnetic resonance imaging scans were used to train three-dimensional U-net models and compare active learning strategies. Uncertainty estimation techniques including Bayesian estimation with dropout, bootstrapping, and margins sampling were compared to random query. Strategies to avoid annotating similar images were also considered. We determined the minimum data necessary to achieve performance equivalent to the model trained on the full dataset (α = 0.05). Bayesian approximation with dropout at training and testing showed results equivalent to that of the full data model (target) with around 30% of the training data needed by random query to achieve target performance (p = 0.018). Annotation redundancy restriction techniques can reduce the training data needed by random query to achieve target performance by 20%. We investigated various active learning strategies to minimize the annotation burden for three-dimensional brain tumor segmentation. Dropout uncertainty estimation achieved target performance with the least annotated data.

Whole-lesion assessment of volume and signal changes after sclerotherapy of extremity venous malformations.

PURPOSE: To investigate quantitative changes in MRI signal intensity (SI) and lesion volume that indicate treatment response and correlate these changes with clinical outcomes after percutaneous sclerotherapy (PS) of extremity venous malformations (VMs). METHODS: VMs were segmented manually on pre- and post-treatment T2-weighted MRI using 3D Slicer to assess changes in lesion volume and SI. Clinical outcomes were scored on a 7-point Likert scale according to patient perception of symptom improvement; treatment response (success or failure) was determined accordingly. RESULTS: Eighty-one patients with VMs underwent 125 PS sessions. Treatment success occurred in 77 patients (95 %). Mean (±SD) changes were -7.9 ± 24 cm3 in lesion volume and -123 ± 162 in SI (both, P <.001). Mean reduction in lesion volume was greater in the success group (-9.4 ± 24 cm3) than in the failure group (21 ± 20 cm3) (P =.006). Overall, lesion volume correlated with treatment response (ρ = -0.3, P =.004). On subgroup analysis, volume change correlated with clinical outcomes in children (ρ = -0.3, P =.03), in sodium tetradecyl sulfate-treated lesions (ρ = -0.5, P =.02), and in foot lesions (ρ = -0.6, P =.04). SI change correlated with clinical outcomes in VMs treated in 1 PS session (ρ = -0.3, P =.01) and in bleomycin-treated lesions (ρ = -0.4, P =.04). CONCLUSIONS: Change in lesion volume is a reliable indicator of treatment response. Lesion volume and SI correlate with clinical outcomes in specific subgroups.

Rigid alignment method for secondary analyses of optical coherence tomography volumes.

Optical coherence tomography (OCT) provides micron level resolution of retinal tissue and is widely used in ophthalmology. Millions of pre-existing OCT images are available from research and clinical databases. Analysis of this data often requires or can benefit significantly from image registration and reduction of speckle noise. One method of reducing noise is to align and average multiple OCT scans together. We propose to use surface feature information and whole volume information to create a novel and simple pipeline that can rigidly align, and average multiple previously acquired 3D OCT volumes from a commercially available OCT device. This pipeline significantly improves both image quality and visualization of clinically relevant image features over single, unaligned volumes from the commercial scanner.

Automated identification of fleck lesions in Stargardt disease using deep learning enhances lesion detection sensitivity and enables morphometric analysis of flecks.

PURPOSE: To classify fleck lesions and assess artificial intelligence (AI) in identifying flecks in Stargardt disease (STGD). METHODS: A retrospective study of 170 eyes from 85 consecutive patients with confirmed STGD. Fundus autofluorescence images were extracted, and flecks were manually outlined. A deep learning model was trained, and a hold-out testing subset was used to compare with manually identified flecks and for graders to assess. Flecks were clustered using K-means clustering. RESULTS: Of the 85 subjects, 45 were female, and the median age was 37 years (IQR 25-59). A subset of subjects (n=41) had clearly identifiable fleck lesions, and an AI was successfully trained to identify these lesions (average Dice score of 0.53, n=18). The AI segmentation had smaller (0.018 compared with 0.034 mm2, p<0.001) but more numerous flecks (75.5 per retina compared with 40.0, p<0.001), but the total size of flecks was not different. The AI model had higher sensitivity to detect flecks but resulted in more false positives. There were two clusters of flecks based on morphology: broadly, one cluster of small round flecks and another of large amorphous flecks. The per cent frequency of small round flecks negatively correlated with subject age (r=-0.31, p<0.005). CONCLUSIONS: AI-based detection of flecks shows greater sensitivity than human graders but with a higher false-positive rate. With further optimisation to address current shortcomings, this approach could be used to prescreen subjects for clinical research. The feasibility and utility of quantifying fleck morphology in conjunction with AI-based segmentation as a biomarker of progression require further study.

Risk of Fractures and Falls in Men with Advanced or Metastatic Prostate Cancer Receiving Androgen Deprivation Therapy and Treated with Novel Androgen Receptor Signalling Inhibitors: A Systematic Review and Meta-analysis of Randomised Controlled Trials.

CONTEXT: The addition of androgen receptor signalling inhibitors (ARSIs) to standard androgen deprivation therapy (ADT) has improved survival outcomes in patients with advanced prostate cancer (PCa). Advanced PCa patients have a higher incidence of osteoporosis, compounded by rapid bone density loss upon commencement of ADT resulting in an increased fracture risk. The effect of treatment intensification with ARSIs on fall and fracture risk is unclear. OBJECTIVE: To assess the risk of falls and fractures in men with PCa treated with ARSIs. EVIDENCE ACQUISITION: A systematic review of EMBASE, MEDLINE, The Cochrane Library, and The Health Technology Assessment Database for randomised control trials between 1990 and June 2023 was conducted in accordance with Preferred Reporting Items for Systematic Review and Meta-analyses guidance. Risk ratios were estimated for the incidence of fracture and fall events. Subgroup analyses by grade of event and disease state were conducted. EVIDENCE SYNTHESIS: Twenty-three studies were eligible for inclusion. Fracture outcomes were reported in 17 studies (N = 18 811) and fall outcomes in 16 studies (N = 16 537). A pooled analysis demonstrated that ARSIs increased the risk of fractures (relative risk [RR] 2.32, 95% confidence interval [CI] 2.00-2.71; p < 0.01) and falls (RR 2.22, 95% CI 1.81-2.72; p < 0.01) compared with control. A subgroup analysis demonstrated an increased risk of both fractures (RR 2.13, 95% CI 1.70-2.67; p < 0.01) and falls (RR 2.19, 95% CI 1.53-3.12; p < 0.0001) in metastatic hormone-sensitive PCa patients, and an increased risk of fractures in the nonmetastatic (RR 2.27, 95% CI 1.60-3.20; p < 0.00001) and metastatic castrate-resistant (RR 2.85, 95% CI 2.16-3.76; p < 0.00001) settings. The key limitations include an inability to distinguish fragility from pathological fractures and potential for a competing risk bias. CONCLUSIONS: Addition of an ARSI to standard ADT significantly increases the risk of fractures and falls in men with prostate cancer. PATIENT SUMMARY: We found a significantly increased risk of both fractures and falls with a combination of novel androgen signalling inhibitors and traditional forms of hormone therapy.

Magnetic Resonance Imaging-Based Assessment of Pancreatic Fat Strongly Correlates With Histology-Based Assessment of Pancreas Composition.

OBJECTIVE: The aim of the study is to assess the relationship between magnetic resonance imaging (MRI)-based estimation of pancreatic fat and histology-based measurement of pancreatic composition. MATERIALS AND METHODS: In this retrospective study, MRI was used to noninvasively estimate pancreatic fat content in preoperative images from high-risk individuals and disease controls having normal pancreata. A deep learning algorithm was used to label 11 tissue components at micron resolution in subsequent pancreatectomy histology. A linear model was used to determine correlation between histologic tissue composition and MRI fat estimation. RESULTS: Twenty-seven patients (mean age 64.0 ± 12.0 years [standard deviation], 15 women) were evaluated. The fat content measured by MRI ranged from 0% to 36.9%. Intrapancreatic histologic tissue fat content ranged from 0.8% to 38.3%. MRI pancreatic fat estimation positively correlated with microanatomical composition of fat (r = 0.90, 0.83 to 0.95], P < 0.001); as well as with pancreatic cancer precursor ( r = 0.65, P < 0.001); and collagen ( r = 0.46, P < 0.001) content, and negatively correlated with pancreatic acinar ( r = -0.85, P < 0.001) content. CONCLUSIONS: Pancreatic fat content, measurable by MRI, correlates to acinar content, stromal content (fibrosis), and presence of neoplastic precursors of cancer.

Vision transformers: The next frontier for deep learning-based ophthalmic image analysis.

Deep learning is the state-of-the-art machine learning technique for ophthalmic image analysis, and convolutional neural networks (CNNs) are the most commonly utilized approach. Recently, vision transformers (ViTs) have emerged as a promising approach, one that is even more powerful than CNNs. In this focused review, we summarized studies that applied ViT-based models to analyze color fundus photographs and optical coherence tomography images. Overall, ViT-based models showed robust performances in the grading of diabetic retinopathy and glaucoma detection. While some studies demonstrated that ViTs were superior to CNNs in certain contexts of use, it is unclear how widespread ViTs will be adopted for ophthalmic image analysis, since ViTs typically require even more training data as compared to CNNs. The studies included were identified from the PubMed and Google Scholar databases using keywords relevant to this review. Only original investigations through March 2023 were included.

Comparative analysis of alignment algorithms for macular optical coherence tomography imaging.

BACKGROUND: Optical coherence tomography (OCT) is the most important and commonly utilized imaging modality in ophthalmology and is especially crucial for the diagnosis and management of macular diseases. Each OCT volume is typically only available as a series of cross-sectional images (B-scans) that are accessible through proprietary software programs which accompany the OCT machines. To maximize the potential of OCT imaging for machine learning purposes, each OCT image should be analyzed en bloc as a 3D volume, which requires aligning all the cross-sectional images within a particular volume. METHODS: A dataset of OCT B-scans obtained from 48 age-related macular degeneration (AMD) patients and 50 normal controls was used to evaluate five registration algorithms. After alignment of B-scans from each patient, an en face surface map was created to measure the registration quality, based on an automatically generated Laplace difference of the surface map-the smoother the surface map, the smaller the average Laplace difference. To demonstrate the usefulness of B-scan alignment, we trained a 3D convolutional neural network (CNN) to detect age-related macular degeneration (AMD) on OCT images and compared the performance of the model with and without B-scan alignment. RESULTS: The mean Laplace difference of the surface map before registration was 27 ± 4.2 pixels for the AMD group and 26.6 ± 4 pixels for the control group. After alignment, the smoothness of the surface map was improved, with a mean Laplace difference of 5.5 ± 2.7 pixels for Advanced Normalization Tools Symmetric image Normalization (ANTs-SyN) registration algorithm in the AMD group and a mean Laplace difference of 4.3 ± 1.4.2 pixels for ANTs in the control group. Our 3D CNN achieved superior performance in detecting AMD, when aligned OCT B-scans were used (AUC 0.95 aligned vs. 0.89 unaligned). CONCLUSIONS: We introduced a novel metric to quantify OCT B-scan alignment and compared the effectiveness of five alignment algorithms. We confirmed that alignment could be improved in a statistically significant manner with readily available alignment algorithms that are available to the public, and the ANTs algorithm provided the most robust performance overall. We further demonstrated that alignment of OCT B-scans will likely be useful for training 3D CNN models.

Brain Tumor Segmentation for Multi-Modal MRI with Missing Information.

Deep convolutional neural networks (DCNNs) have shown promise in brain tumor segmentation from multi-modal MRI sequences, accommodating heterogeneity in tumor shape and appearance. The fusion of multiple MRI sequences allows networks to explore complementary tumor information for segmentation. However, developing a network that maintains clinical relevance in situations where certain MRI sequence(s) might be unavailable or unusual poses a significant challenge. While one solution is to train multiple models with different MRI sequence combinations, it is impractical to train every model from all possible sequence combinations. In this paper, we propose a DCNN-based brain tumor segmentation framework incorporating a novel sequence dropout technique in which networks are trained to be robust to missing MRI sequences while employing all other available sequences. Experiments were performed on the RSNA-ASNR-MICCAI BraTS 2021 Challenge dataset. When all MRI sequences were available, there were no significant differences in performance of the model with and without dropout for enhanced tumor (ET), tumor (TC), and whole tumor (WT) (p-values 1.000, 1.000, 0.799, respectively), demonstrating that the addition of dropout improves robustness without hindering overall performance. When key sequences were unavailable, the network with sequence dropout performed significantly better. For example, when tested on only T1, T2, and FLAIR sequences together, DSC for ET, TC, and WT increased from 0.143 to 0.486, 0.431 to 0.680, and 0.854 to 0.901, respectively. Sequence dropout represents a relatively simple yet effective approach for brain tumor segmentation with missing MRI sequences.

Abiraterone acetate plus prednisolone with or without enzalutamide for patients with metastatic prostate cancer starting androgen deprivation therapy: final results from two randomised phase 3 trials of the STAMPEDE platform protocol.

BACKGROUND: Abiraterone acetate plus prednisolone (herein referred to as abiraterone) or enzalutamide added at the start of androgen deprivation therapy improves outcomes for patients with metastatic prostate cancer. Here, we aimed to evaluate long-term outcomes and test whether combining enzalutamide with abiraterone and androgen deprivation therapy improves survival. METHODS: We analysed two open-label, randomised, controlled, phase 3 trials of the STAMPEDE platform protocol, with no overlapping controls, conducted at 117 sites in the UK and Switzerland. Eligible patients (no age restriction) had metastatic, histologically-confirmed prostate adenocarcinoma; a WHO performance status of 0-2; and adequate haematological, renal, and liver function. Patients were randomly assigned (1:1) using a computerised algorithm and a minimisation technique to either standard of care (androgen deprivation therapy; docetaxel 75 mg/m2 intravenously for six cycles with prednisolone 10 mg orally once per day allowed from Dec 17, 2015) or standard of care plus abiraterone acetate 1000 mg and prednisolone 5 mg (in the abiraterone trial) orally or abiraterone acetate and prednisolone plus enzalutamide 160 mg orally once a day (in the abiraterone and enzalutamide trial). Patients were stratified by centre, age, WHO performance status, type of androgen deprivation therapy, use of aspirin or non-steroidal anti-inflammatory drugs, pelvic nodal status, planned radiotherapy, and planned docetaxel use. The primary outcome was overall survival assessed in the intention-to-treat population. Safety was assessed in all patients who started treatment. A fixed-effects meta-analysis of individual patient data was used to compare differences in survival between the two trials. STAMPEDE is registered with ClinicalTrials.gov (NCT00268476) and ISRCTN (ISRCTN78818544). FINDINGS: Between Nov 15, 2011, and Jan 17, 2014, 1003 patients were randomly assigned to standard of care (n=502) or standard of care plus abiraterone (n=501) in the abiraterone trial. Between July 29, 2014, and March 31, 2016, 916 patients were randomly assigned to standard of care (n=454) or standard of care plus abiraterone and enzalutamide (n=462) in the abiraterone and enzalutamide trial. Median follow-up was 96 months (IQR 86-107) in the abiraterone trial and 72 months (61-74) in the abiraterone and enzalutamide trial. In the abiraterone trial, median overall survival was 76·6 months (95% CI 67·8-86·9) in the abiraterone group versus 45·7 months (41·6-52·0) in the standard of care group (hazard ratio [HR] 0·62 [95% CI 0·53-0·73]; p<0·0001). In the abiraterone and enzalutamide trial, median overall survival was 73·1 months (61·9-81·3) in the abiraterone and enzalutamide group versus 51·8 months (45·3-59·0) in the standard of care group (HR 0·65 [0·55-0·77]; p<0·0001). We found no difference in the treatment effect between these two trials (interaction HR 1·05 [0·83-1·32]; pinteraction=0·71) or between-trial heterogeneity (I2 p=0·70). In the first 5 years of treatment, grade 3-5 toxic effects were higher when abiraterone was added to standard of care (271 [54%] of 498 vs 192 [38%] of 502 with standard of care) and the highest toxic effects were seen when abiraterone and enzalutamide were added to standard of care (302 [68%] of 445 vs 204 [45%] of 454 with standard of care). Cardiac causes were the most common cause of death due to adverse events (five [1%] with standard of care plus abiraterone and enzalutamide [two attributed to treatment] and one (<1%) with standard of care in the abiraterone trial). INTERPRETATION: Enzalutamide and abiraterone should not be combined for patients with prostate cancer starting long-term androgen deprivation therapy. Clinically important improvements in survival from addition of abiraterone to androgen deprivation therapy are maintained for longer than 7 years. FUNDING: Cancer Research UK, UK Medical Research Council, Swiss Group for Clinical Cancer Research, Janssen, and Astellas.

The effect of neonatal vaccination for bovine respiratory disease in the face of a dual challenge with bovine viral diarrhea virus and Mannheimia hemolytica.

Bovine respiratory disease is the greatest threat to calf health. In this study, colostrum-fed dairy X beef calves were vaccinated at ∼30 days of age with an adjuvanted parenteral vaccine containing modified live bovine viral diarrhea virus (BVDV) type 1 and type 2, bovine herpesvirus 1 (BHV-1), bovine parainfluenza type 3 virus (PI3V) and bovine respiratory syncytial virus (BRSV) andM. haemolyticatoxoid (Group 1), or intranasal temperature-sensitive BHV-1, BRSV and PI3V concurrently witha parenteral vaccine containing modified live BVDV type 1 and type 2 andM. haemolyticatoxoid (Group 2) or a placebo (Group 3). The calves were challenged ∼150 days post vaccination intranasally with BVDV 1b and then 7 days later intratracheally withM. haemolytica. The calves wereeuthanized 6 days after theM. haemolyticachallenge. Clinical signs following BVDV infection were similar in all groups. There was increased rectal temperatures in the Groups 2 and 3 on day 3 and in Group 3 on days 8-13. Group 1 animals had a slight leukopenia following BVDV infection while Groups 2 and 3 had greater leukopenia. BVDV type 1 and 2 serum titers increased in Group 1 following vaccination while these titers waned in Groups 2 and 3. There were higher levels of BVDV in the buffy coats and nasal samples in Group 2 and Group 3 versus Group 1 (p < 0.01). Interferon-gamma response was higher (p < 0.01) in Group 1 animals than Groups 2 and 3. Group 1 had the lowest percent pneumonic tissue (1.6%) while Group 2 vaccinates had 3.7% and the control Group 3 was 5.3%. Vaccination in the face of maternal antibody with a parenteral adjuvanted vaccine resulted in better protection than the regimen of an intranasal vaccine anda parenteral adjuvanted BVDV andM haemolyticacombination vaccine in a BVDV-M. haemolyticadual challenge.

Combining physics-based models with deep learning image synthesis and uncertainty in intraoperative cone-beam CT of the brain.

BACKGROUND: Image-guided neurosurgery requires high localization and registration accuracy to enable effective treatment and avoid complications. However, accurate neuronavigation based on preoperative magnetic resonance (MR) or computed tomography (CT) images is challenged by brain deformation occurring during the surgical intervention. PURPOSE: To facilitate intraoperative visualization of brain tissues and deformable registration with preoperative images, a 3D deep learning (DL) reconstruction framework (termed DL-Recon) was proposed for improved intraoperative cone-beam CT (CBCT) image quality. METHODS: The DL-Recon framework combines physics-based models with deep learning CT synthesis and leverages uncertainty information to promote robustness to unseen features. A 3D generative adversarial network (GAN) with a conditional loss function modulated by aleatoric uncertainty was developed for CBCT-to-CT synthesis. Epistemic uncertainty of the synthesis model was estimated via Monte Carlo (MC) dropout. Using spatially varying weights derived from epistemic uncertainty, the DL-Recon image combines the synthetic CT with an artifact-corrected filtered back-projection (FBP) reconstruction. In regions of high epistemic uncertainty, DL-Recon includes greater contribution from the FBP image. Twenty paired real CT and simulated CBCT images of the head were used for network training and validation, and experiments evaluated the performance of DL-Recon on CBCT images containing simulated and real brain lesions not present in the training data. Performance among learning- and physics-based methods was quantified in terms of structural similarity (SSIM) of the resulting image to diagnostic CT and Dice similarity metric (DSC) in lesion segmentation compared to ground truth. A pilot study was conducted involving seven subjects with CBCT images acquired during neurosurgery to assess the feasibility of DL-Recon in clinical data. RESULTS: CBCT images reconstructed via FBP with physics-based corrections exhibited the usual challenges to soft-tissue contrast resolution due to image non-uniformity, noise, and residual artifacts. GAN synthesis improved image uniformity and soft-tissue visibility but was subject to error in the shape and contrast of simulated lesions that were unseen in training. Incorporation of aleatoric uncertainty in synthesis loss improved estimation of epistemic uncertainty, with variable brain structures and unseen lesions exhibiting higher epistemic uncertainty. The DL-Recon approach mitigated synthesis errors while maintaining improvement in image quality, yielding 15%-22% increase in SSIM (image appearance compared to diagnostic CT) and up to 25% increase in DSC in lesion segmentation compared to FBP. Clear gains in visual image quality were also observed in real brain lesions and in clinical CBCT images. CONCLUSIONS: DL-Recon leveraged uncertainty estimation to combine the strengths of DL and physics-based reconstruction and demonstrated substantial improvements in the accuracy and quality of intraoperative CBCT. The improved soft-tissue contrast resolution could facilitate visualization of brain structures and support deformable registration with preoperative images, further extending the utility of intraoperative CBCT in image-guided neurosurgery.

Interrater Reliability of Cervical Neural Foraminal Stenosis Using Traditional and Splayed Reconstructions: Analysis of One Hundred Scans.

OBJECTIVE: The oblique sagittal orientation of the cervical neural foramina hinders the evaluation of cervical neural foraminal stenosis (CNFS) on traditional axial and sagittal slices. Traditional image reconstruction techniques to generate oblique slices provide only a view of the foramina unilaterally. We present a simple technique for generating splayed slices that show the bilateral neuroforamina simultaneously and assess its reliability compared with traditional axial windows. METHODS: Cervical computed tomography (CT) scans from 100 patients were retrospectively collected and de-identified. The axial slices were reformatted into a curved reformat with the plane of the reformat extending across the bilateral neuroforamina. The foramina along the C2-T1 vertebral levels were assessed by 4 neuroradiologists using the axial and splayed slices. The intrarater agreement across the axial and splayed slices for a given foramen and the interrater agreement for the axial and splayed slices individually were calculated using the Cohen κ statistic. RESULTS: Interrater agreement was overall higher for the splayed slices (κ = 0.25) compared with the axial slices (κ = 0.20). The splayed slices were more likely to have fair agreement across raters compared with the axial slices. Intrarater agreement between the axial and splayed slices was poorer for residents compared with fellows. CONCLUSIONS: Splayed reconstructions showing the bilateral neuroforamina en face can be readily generated from axial CT imaging. These splayed reconstructions can improve the consistency of CNFS evaluation compared with traditional CT slices and should be considered in the workup of CNFS, particularly for less experienced readers.

Prediction of visual impairment in retinitis pigmentosa using deep learning and multimodal fundus images.

BACKGROUND: The efficiency of clinical trials for retinitis pigmentosa (RP) treatment is limited by the screening burden and lack of reliable surrogate markers for functional end points. Automated methods to determine visual acuity (VA) may help address these challenges. We aimed to determine if VA could be estimated using confocal scanning laser ophthalmoscopy (cSLO) imaging and deep learning (DL). METHODS: Snellen corrected VA and cSLO imaging were obtained retrospectively. The Johns Hopkins University (JHU) dataset was used for 10-fold cross-validations and internal testing. The Amsterdam University Medical Centers (AUMC) dataset was used for external independent testing. Both datasets had the same exclusion criteria: visually significant media opacities and images not centred on the central macula. The JHU dataset included patients with RP with and without molecular confirmation. The AUMC dataset only included molecularly confirmed patients with RP. Using transfer learning, three versions of the ResNet-152 neural network were trained: infrared (IR), optical coherence tomography (OCT) and combined image (CI). RESULTS: In internal testing (JHU dataset, 2569 images, 462 eyes, 231 patients), the area under the curve (AUC) for the binary classification task of distinguishing between Snellen VA 20/40 or better and worse than Snellen VA 20/40 was 0.83, 0.87 and 0.85 for IR, OCT and CI, respectively. In external testing (AUMC dataset, 349 images, 166 eyes, 83 patients), the AUC was 0.78, 0.87 and 0.85 for IR, OCT and CI, respectively. CONCLUSIONS: Our algorithm showed robust performance in predicting visual impairment in patients with RP, thus providing proof-of-concept for predicting structure-function correlation based solely on cSLO imaging in patients with RP.

Active shape model registration of ocular structures in computed tomography images.

Purpose. The goal of this work is to create an active shape model segmentation method based on the statistical shape model of five regions of the globe on computed tomography (CT) scans and to use the method to categorize normal globe from globe injury.Methods. A set of 78 normal globes imaged with CT scans were manually segmented (vitreous cavity, lens, sclera, anterior chamber, and cornea) by two graders. A statistical shape model was created from the regions. An active shape model was trained using the manual segmentations and the statistical shape model and was assessed using leave-one-out cross validations. The active shape model was then applied to a set of globes with open globe injures, and the segmentations were compared to those of normal globes, in terms of the standard deviations away from normal.Results. The active shape model (ASM) segmentation compared well to ground truth, based on Dice similarity coefficient score in a leave-one-out experiment: 90.2% ± 2.1% for the cornea, 92.5% ± 3.5% for the sclera, 87.4% ± 3.7% for the vitreous cavity, 83.5% ± 2.3% for the anterior chamber, and 91.2% ± 2.4% for the lens. A preliminary set of CT scans of patients with open globe injury were segmented using the ASM and the shape of each region was quantified. The sclera and vitreous cavity were statistically different in shape from the normal. The Zone 1 and Zone 2 globes were statistically different than normal from the cornea and anterior chamber. Both results are consistent with the definition of the zonal injuries in OGI.Conclusion. The ASM results were found to be reproducible and accurately correlated with manual segmentations. The quantitative metrics derived from ASM of globes with OGI are consistent with existing medical knowledge in terms of structural deformation.

Arterial Spin Labeling Imaging Characteristics of Anti-leucine-rich Glioma-Inactivated 1 Encephalitis: A Qualitative and Quantitative Analysis.

Background and Significance: Autoimmune encephalitis (AE) is a rare group of diseases that can present with stroke-like symptoms. Anti-leucine-rich glioma inactivated 1 (LGI1) encephalitis is an AE subtype that is infrequently associated with neoplasms and highly responsive to prompt immunotherapy treatment. Therefore, accurate diagnosis of LGI1 AE is essential in timely patient management. Neuroimaging plays a critical role in evaluating stroke and stroke mimics such as AE. Arterial Spin Labeling (ASL) is an MRI perfusion modality that measures cerebral blood flow (CBF) and is increasingly used in everyday clinical practice for stroke and stroke mimic assessment as a non-contrast sequence. Our goal in this preliminary study is to demonstrate the added value of ASL in detecting LGI1 AE for prompt diagnosis and treatment. Methods: In this retrospective single center study, we identified six patients with seropositive LGI1 AE who underwent baseline MRI with single delay 3D pseudocontinuous ASL (pCASL), including five males and one female between ages 28 and 76 years, with mean age of 55 years. Two neuroradiologists qualitatively interpreted the ASL images by visual inspection of CBF using a two-point scale (increased, decreased) when compared to both the ipsilateral and contralateral unaffected temporal and non-temporal cortex. The primary measures on baseline ASL evaluation were a) presence of ASL signal abnormality, b) if present, signal characterization based on the two-point scale, c) territorial vascular distribution, d) localization, and e) laterality. Quantitative assessment was also performed on postprocessed pCASL cerebral blood flow (CBF) maps. The obtained CBF values were then compared between the affected temporal cortex and each of the unaffected ipsilateral parietal, contralateral temporal, and contralateral parietal cortices. Results: On consensus qualitative assessment, all six patients demonstrated ASL hyperperfusion and corresponding FLAIR hyperintensity in the hippocampus and/or amygdala in a non-territorial distribution (6/6, 100%). The ASL hyperperfusion was found in the right hippocampus or amygdala in 5/6 (83%) of cases. Four of the six patients underwent initial follow-up imaging where all four showed resolution of the initial ASL hyperperfusion. In the same study on structural imaging, all four patients were also diagnosed with mesial temporal sclerosis (MTS). Quantitative assessment was separately performed and demonstrated markedly increased CBF values in the affected temporal cortex (mean, 111.2 ml/min/100 g) compared to the unaffected ipsilateral parietal cortex (mean, 49 ml/min/100 g), contralateral temporal cortex (mean, 58.2 ml/min/100 g), and contralateral parietal cortex (mean, 52.2 ml/min/100 g). Discussion: In this preliminary study of six patients, we demonstrate an ASL hyperperfusion pattern, with a possible predilection for the right mesial temporal lobe on both qualitative and quantitative assessments in patients with seropositive LGI1. Larger scale studies are necessary to further characterize the strength of these associations.

Deformable 3D-2D image registration and analysis of global spinal alignment in long-length intraoperative spine imaging.

BACKGROUND: Spinal deformation during surgical intervention (caused by patient positioning and/or the correction of malalignment) confounds conventional navigation due to the assumptions of rigid transformation. Moreover, the ability to accurately quantify spinal alignment in the operating room would provide an assessment of the surgical product via metrics that correlate with clinical outcomes. PURPOSE: A method for deformable 3D-2D registration of preoperative CT to intraoperative long-length tomosynthesis images is reported for an accurate 3D evaluation of device placement in the presence of spinal deformation and automated evaluation of global spinal alignment (GSA). METHODS: Long-length tomosynthesis ("Long Film," LF) images were acquired using an O-arm imaging system (Medtronic, Minneapolis USA). A deformable 3D-2D patient registration was developed using multi-scale masking (proceeding from the full-length image to local subvolumes about each vertebra) to transform vertebral labels and planning information from preoperative CT to the LF images. Automatic measurement of GSA (main thoracic kyphosis [MThK] and lumbar lordosis [LL]) was obtained using a spline fit to registered labels. The "Known-Component Registration" method for device registration was adapted to the multi-scale process for 3D device localization from orthogonal LF images. The multi-scale framework was evaluated using a deformable spine phantom in which pedicle screws were inserted, and deformations were induced over a range in LL ∼25°-80°. Further validation was carried out in a cadaver study with implanted pedicle screws and a similar range of spinal deformation. The accuracy of patient and device registration was evaluated in terms of 3D translational error and target registration error, respectively, and the accuracies of automatic GSA measurements were compared to manual annotation. RESULTS: Phantom studies demonstrated accurate registration via the multi-scale framework for all vertebral levels in both the neutral and deformed spine: median (interquartile range, IQR) patient registration error was 1.1 mm (0.7-1.9 mm IQR). Automatic measures of MThK and LL agreed with manual delineation within -1.1° ± 2.2° and 0.7° ± 2.0° (mean and standard deviation), respectively. Device registration error was 0.7 mm (0.4-1.0 mm IQR) at the screw tip and 0.9° (1.0°-1.5°) about the screw trajectory. Deformable 3D-2D registration significantly outperformed conventional rigid registration (p < 0.05), which exhibited device registration errors of 2.1 mm (0.8-4.1 mm) and 4.1° (1.2°-9.5°). Cadaver studies verified performance under realistic conditions, demonstrating patient registration error of 1.6 mm (0.9-2.1 mm); MThK within -4.2° ± 6.8° and LL within 1.7° ± 3.5°; and device registration error of 0.8 mm (0.5-1.9 mm) and 0.7° (0.4°-1.2°) for the multi-scale deformable method, compared to 2.5 mm (1.0-7.9 mm) and 2.3° (1.6°-8.1°) for rigid registration (p < 0.05). CONCLUSION: The deformable 3D-2D registration framework leverages long-length intraoperative imaging to achieve accurate patient and device registration over the extended lengths of the spine (up to 64 cm) even with strong anatomical deformation. The method offers a new means for the quantitative validation of spinal correction (intraoperative GSA measurement) and the 3D verification of device placement in comparison to preoperative images and planning data.