Manual Segmentation Techniques Research Articles

The effect of imaging programs and segmentation methods on the accuracy of volume measurements of teeth

ObjectiveThis study investigated whether there are differences between software programs, voxel sizes, segmentation techniques, and intraoral scanners in terms of volume measurement in incisor teeth. Study DesignThirty extracted teeth were scanned using a 3D intraoral scanner. Physical volumes were measured using the water displacement method (WDM) as the gold standard. Cone Beam Computed Tomography (CBCT) images were taken in two voxel sizes (0.3- 0.4 voxel). The volumes of the teeth were calculated using the manual segmentation technique in the 3D Doctor and ImageJ (Fiji) Program, manual and automatic segmentation methods in the ITK-Snap program, and automatic segmentation methods in the 3D Slicer program. The data were analyzed using the Statistical Package for Social Sciences (SPSS, v 20.0). ResultsThere is a significant difference between WDM and 3D Doctor volume measurements at the voxel size of 0.4 (p <0.05). The 3D Doctor program showed more than 13% difference compared to the WDM. There is no significant difference between WDM, intraoral scanner, ITK-Snap, ImageJ (Fiji), 3D Slicer, and 3D-Doctor volume measurements at the voxel size of 0.3. ConclusionMeasurements with manual segmentation in ITK-Snap and ImageJ programs give the closest results to the physical volume measurements of the teeth. The automatic segmentation method in ITK-Snap and 3D Slicer programs may be preferred due to its ease of use and less time consuming. In the 3D Doctor software, volume measurements tend to increase with larger voxel size. Statement of Clinical RelevanceDifferent software programs, voxel size, and segmentation techniques used in CBCT images affect the measurement results of tooth volumes. Very close to physical volume was obtained with the intraoral scanner. Measurements obtained through manual segmentation yield more accurate results.

Automation of green roof plant cover measurements using machine learning and a comparison of digital and thermal imaging techniques

AbstractAimsPost‐analyses of digital red, green, blue (RGB) and thermal images have become increasingly popular as modern approaches to plant cover analysis. Image analyses are often coupled with semi‐automated or automated workflows to reduce the amount of human labor input compared with traditional manual procedures. This study aims to evaluate and compare different image segmentation methods for plant cover analysis using digital RGB and thermal images, focusing on the effectiveness of semi‐automated and manual segmentation techniques in monitoring plant cover on green roofs.LocationAn Extensive green roof in the City of Toronto.MethodsWe surveyed the plant cover of an extensive green roof using digital and thermal imagery. The plant cover values were obtained using three methods: traditional manual segmentation based on a visual examination (MS), ImageJ Color Threshold (CT) and Trainable Weka Segmentation (TWS), all performed within FIJI (a distribution of ImageJ). Manual segmentation based on visual examination was used as a reference standard.ResultsSignificant correlation was found between the cover estimation using the CT and TWS methods relative to MS, and between cover estimation using the thermal image and the RGB image. TWS overestimated plant cover on thermal images while producing an underestimation on RGB images. CT demonstrated a performance closer to MS than TWS, indicating that manually customized methods produced results more aligned with MS. The estimated cover values by MS were not significantly affected by the image type (digital RGB or thermal).ConclusionsResults suggest that RGB and thermal imaging techniques may provide complementary results and reveal unique information regarding the functioning of green roofs. The accuracy of supervised machine‐learning methods could be enhanced with site‐specific data to provide a more accurate and efficient estimation of plant cover, which might be beneficial for long‐term studies on green roofs and ecological sites in remote locations.

ViNe-Seg: deep-learning-assisted segmentation of visible neurons and subsequent analysis embedded in a graphical user interface.

Segmentation of neural somata is a crucial and usually the most time-consuming step in the analysis of optical functional imaging of neuronal microcircuits. In recent years, multiple auto-segmentation tools have been developed to improve the speed and consistency of the segmentation process, mostly, using deep learning approaches. Current segmentation tools, while advanced, still encounter challenges in producing accurate segmentation results, especially in datasets with a low signal-to-noise ratio. This has led to a reliance on manual segmentation techniques. However, manual methods, while customized to specific laboratory protocols, can introduce variability due to individual differences in interpretation, potentially affecting dataset consistency across studies. In response to this challenge, we present ViNe-Seg: a deep-learning-based semi-automatic segmentation tool that offers (i) detection of visible neurons, irrespective of their activity status; (ii) the ability to perform segmentation during an ongoing experiment; (iii) a user-friendly graphical interface that facilitates expert supervision, ensuring precise identification of Regions of Interest; (iv) an array of segmentation models with the option of training custom models and sharing them with the community; and (v) seamless integration of subsequent analysis steps. ViNe-Seg code and documentation are publicly available at https://github.com/NiRuff/ViNe-Seg and can be installed from https://pypi.org/project/ViNeSeg/.

U-Net-Based Automatic Segmentation of Sphenoid Sinus Fluid in Drowning Cases Using Postmortem CT Images: A Feasibility Study

Detecting sphenoid sinus fluid (SSF) is an additional finding in autopsies for diagnosing drowning. SSF can provide additional forensic evidence through laboratory tests such as diatom and electrolyte analyses. If drowning is suspected, accurately assessing the presence and volume of SSF during an autopsy is crucial. Utilizing postmortem computed tomography (PMCT) images could aid in accurately sampling SSF. Accurately segmenting the region of interest is essential for volume analysis using computed tomography images. However, manual segmentation techniques are labor-intensive and time-consuming, and their success depends on the experience of the observer. Therefore, this study aimed to develop a U-Net–based deep learning model for the automatic segmentation of SSF in drowning cases using PMCT images and to evaluate the performance of the model. We retrospectively reviewed 34 drowning cases in which both PMCT scans and forensic autopsies were performed at our institution. The U-Net architecture of deep learning was used for automatic segmentation. The proposed model achieved the Dice similarity coefficient (DSC) and Intersection over Union (IoU) of a maximum of 95.85% and 92.03%, a minimum of 0% and 0%, and an average of 77.15% and 67.18%, respectively. Although the average DSC and IoU did not show high similarity, this study showed that PMCT images can be used for automatic segmentation of SSF in drowning cases, which could improve the performance with sufficient dataset acquisition and further model training.

Radiological measurements of lacrimal gland in thyroid eye disease

PurposeLacrimal gland enlargement is a common feature of thyroid eye disease (TED) and has been positively correlated with the clinical activity score. Although lacrimal gland volume is the preferred measure of lacrimal gland size, its calculation is not clinically translatable due to the expertise, time and advanced software required. The aim of our study is to determine whether the lacrimal gland volume in patients with TED undergoing magnetic resonance imaging (MRI) can be estimated using simpler lacrimal gland linear and area measurements.MethodsA retrospective review of 102 orbits (51 patients) with TED who underwent orbital MRI was conducted. The maximum length, width, and area of the lacrimal gland were measured in axial and coronal sections. Lacrimal gland volume was calculated by using a manual segmentation technique on all consecutive axial slices on commercially available software, OsiriX. All quantitative measurements were correlated with the lacrimal gland volume.ResultsMean age of participants was 59 ± 16 years, and 67% (n = 34) were females. With multivariate analyses, combined lacrimal gland axial and coronal areas strongly correlated with volume (r = 0.843, p < 0.01). Strong univariate predictors of volume included axial area (r = 0.704, p < 0.01) and coronal area (r = 0.722, p < 0.01), while moderate predictors included axial length (r = 0.523, p < 0.01), axial width (r = 0.521, p < 0.01), coronal length (r = 0.450, p < 0.01), and coronal width (r = 0.649, p < 0.01).ConclusionIn patients with thyroid eye disease, lacrimal gland volume can be estimated using axial and coronal areas, which is simpler and more time efficient than calculating volumes.

Evaluating the clinical applicability of neural networks for meningioma tumor segmentation on 3D MRI

Magnetic resonance imaging (MRI), in both 2D and 3D configurations, serves as the leading non-invasive and non-ionizing technique for the detection and clinical assessment of brain tumors. 3D multiparametric MRIs (mpMRIs) offer enhanced spatial and biological context over other MRIs, which can help clinicians plan tailored treatment and therapy. Presently, surgical intervention remains the primary treatment method for brain tumors, requiring accurate segmentation of tumor tissue on MRI scans. However, manual tumor segmentation and traditional machine learning techniques are labor intensive and subject to observer bias. Deep learning models can provide enhanced accuracy and precision in tumor segmentation due to their ability to automatically extract intricate features and patterns from MRI scans. In this study, MRI scans from 358 meningioma patients were used to train a series of deep learning models to segment the tumor subregions. We hypothesized that a deep learning framework designed for meningioma tumor segmentation and trained utilizing 3D mpMRIs would demonstrate superior segmentation accuracy compared to a framework trained on 2D MRIs or single-sequence 3D MRIs. We compared the model predictions with the ground truth labels using similarity metrics to assess accuracy and clinical applicability. The model trained on 3D mpMRI scans showed reliable performance for meningioma subregion segmentation with a Dice Similarity Coefficient score of 0.91 and a median Sensitivity of 91.38%. It outperformed the other models due to the additional spatial context provided by 3D mpMRIs, suggesting it could be readily used in clinical practice to help guide treatment strategy.

EXTH-33. RAGE INHIBITOR (TTP488) AS AN ALTERNATIVE TO DEXAMETHASONE FOR MANAGING CEREBRAL EDEMA FOLLOWING BRAIN TUMOR SURGERY

Abstract BACKGROUND Cerebral edema (CE) commonly occurs after neurosurgical procedures and contributes to patient morbidity and neurological deterioration. Corticosteroids alleviate brain edema during the perioperative period, but they impose several undesirable side effects and interfere with immunotherapies in patients with malignant brain tumors. OBJECTIVE To assess the anti-inflammatory effects of TTP488 (inhibitor of receptor for advanced glycation end products) on CE formation following tumor resection in glioma models. METHODS Mice harboring orthotopic CT-2A gliomas were randomly assigned to three groups and underwent tumor resection. TTP488, dexamethasone, and vehicle were administered perioperatively (day -4 pre-op to day 7 post-op) and post-operative neurological function was examined. CE was assessed on serial brain MRIs over a 7-day period, and using manual segmentation techniques, tumor-related volumes were calculated. In a separate survival experiment, the effect of each drug on anti-PD-1 immunotherapy response was measured. RESULTS In each group, CE peaked on post-operative day 2 and gradually diminished by day 7. On post-operative day one, mice treated with TTP488 showed a marked reduction in CE compared to those treated with either dexamethasone or vehicle with mean CE volumes of 1.75±1.08%, 2.69±0.77%, and 3.87±1.07%, respectively (P=0.02, TTP488 vs vehicle). Furthermore, the TTP488 group had better functional recovery scores (P=0.04) and wound healing (dehiscence, P=0.00003) compared to the vehicle group. Interestingly, unlike dexamethasone, TTP488 did not abrogate the efficacy of immunotherapy in the CT-2A glioma model. CONCLUSION TTP488 was as effective as dexamethasone in alleviating CE caused by tumor resection and did not interfere with the efficacy of immunotherapy. TTP488 could potentially be an alternative to dexamethasone in the perioperative management of patients undergoing tumor resection or other neurosurgical interventions.

U-Net Architectures for Prostate Cancer Radiation Therapy

Prostate Cancer (PCa) was first diagnosed in 1853 as cirrhosis of the prostate gland. Silently asymptomatic, PCa is usually diagnosed with Digital Rectal Examination (DRE) and Prostate Specific Antigen (PSA) levels. Since the first treatment of an advanced prostatic malignancy with X-rays by Imbert and Imbert in 1904, External Beam Radiation Therapy (EBRT) is now a curative option for localised and locally advanced disease and a palliative option for the metastatic low-volume disease. With the introduction of computers in EBRT and better imaging techniques, volume delineation is still a very time-consuming task. While Deep Learning methods are urged, EBRT systems still rely on manual or semi-automatic segmentation techniques. The U-Net architecture was specially designed for medical image segmentation presenting promising results. This literature review gathers work using U-Net architectures, outlining methods, techniques and obtained outcomes as a potential foundation for an automated segmentation framework for PCa.

Convolutional neural networks for the automatic segmentation of lumbar paraspinal muscles in people with low back pain

The size, shape, and composition of paraspinal muscles have been widely reported in disorders of the cervical and lumbar spine. Measures of size, shape, and composition have required time-consuming and rater-dependent manual segmentation techniques. Convolutional neural networks (CNNs) provide alternate timesaving, state-of-the-art performance measures, which could realise clinical translation. Here we trained a CNN for the automatic segmentation of lumbar paraspinal muscles and determined the impact of CNN architecture and training choices on segmentation performance. T2-weighted MRI axial images from 76 participants (46 female; age (SD): 45.6 (12.8) years) with low back pain were used to train CNN models to segment the multifidus, erector spinae, and psoas major muscles (left and right segmented separately). Using cross-validation, we compared 2D and 3D CNNs with and without data augmentation. Segmentation accuracy was compared between the models using the Sørensen-Dice index as the primary outcome measure. The effect of increasing network depth on segmentation accuracy was also investigated. Each model showed high segmentation accuracy (Sørensen-Dice index ≥ 0.885) and excellent reliability (ICC2,1 ≥ 0.941). Overall, across all muscles, 2D models performed better than 3D models (p = 0.012), and training without data augmentation outperformed training with data augmentation (p < 0.001). The 2D model trained without data augmentation demonstrated the highest average segmentation accuracy. Increasing network depth did not improve accuracy (p = 0.771). All trained CNN models demonstrated high accuracy and excellent reliability for segmenting lumbar paraspinal muscles. CNNs can be used to efficiently and accurately extract measures of paraspinal muscle health from MRI.

Design and validation of a semi-automatic bone segmentation algorithm from MRI to improve research efficiency

Segmentation of medical images into different tissue types is essential for many advancements in orthopaedic research; however, manual segmentation techniques can be time- and cost-prohibitive. The purpose of this work was to develop a semi-automatic segmentation algorithm that leverages gradients in spatial intensity to isolate the patella bone from magnetic resonance (MR) images of the knee that does not require a training set. The developed algorithm was validated in a sample of four human participants (in vivo) and three porcine stifle joints (ex vivo) using both magnetic resonance imaging (MRI) and computed tomography (CT). We assessed the repeatability (expressed as mean ± standard deviation) of the semi-automatic segmentation technique on: (1) the same MRI scan twice (Dice similarity coefficient = 0.988 ± 0.002; surface distance = − 0.01 ± 0.001 mm), (2) the scan/re-scan repeatability of the segmentation technique (surface distance = − 0.02 ± 0.03 mm), (3) how the semi-automatic segmentation technique compared to manual MRI segmentation (surface distance = − 0.02 ± 0.08 mm), and (4) how the semi-automatic segmentation technique compared when applied to both MRI and CT images of the same specimens (surface distance = − 0.02 ± 0.06 mm). Mean surface distances perpendicular to the cartilage surface were computed between pairs of patellar bone models. Critically, the semi-automatic segmentation algorithm developed in this work reduced segmentation time by approximately 75%. This method is promising for improving research throughput and potentially for use in generating training data for deep learning algorithms.

SuRVoS 2: Accelerating Annotation and Segmentation for Large Volumetric Bioimage Workflows Across Modalities and Scales.

As sample preparation and imaging techniques have expanded and improved to include a variety of options for larger sized and numbers of samples, the bottleneck in volumetric imaging is now data analysis. Annotation and segmentation are both common, yet difficult, data analysis tasks which are required to bring meaning to the volumetric data. The SuRVoS application has been updated and redesigned to provide access to both manual and machine learning-based segmentation and annotation techniques, including support for crowd sourced data. Combining adjacent, similar voxels (supervoxels) provides a mechanism for speeding up segmentation both in the painting of annotation and by training a segmentation model on a small amount of annotation. The support for layers allows multiple datasets to be viewed and annotated together which, for example, enables the use of correlative data (e.g. crowd-sourced annotations or secondary imaging techniques) to guide segmentation. The ability to work with larger data on high-performance servers with GPUs has been added through a client-server architecture and the Pytorch-based image processing and segmentation server is flexible and extensible, and allows the implementation of deep learning-based segmentation modules. The client side has been built around Napari allowing integration of SuRVoS into an ecosystem for open-source image analysis while the server side has been built with cloud computing and extensibility through plugins in mind. Together these improvements to SuRVoS provide a platform for accelerating the annotation and segmentation of volumetric and correlative imaging data across modalities and scales.

Validating a Semi-Automated Technique for Segmenting Femoral Articular Cartilage on Ultrasound Images.

ObjectiveTo validate a semi-automated technique to segment ultrasound-assessed femoral cartilage without compromising segmentation accuracy to a traditional manual segmentation technique in participants with an anterior cruciate ligament injury (ACL).DesignWe recruited 27 participants with a primary unilateral ACL injury at a pre-operative clinic visit. One investigator performed a transverse suprapatellar ultrasound scan with the participant’s ACL injured knee in maximum flexion. Three femoral cartilage ultrasound images were recorded. A single expert reader manually segmented the femoral cartilage cross-sectional area in each image. In addition, we created a semi-automatic program to segment the cartilage using a random walker-based method. We quantified the average cartilage thickness and echo-intensity for the manual and semi-automated segmentations. Intraclass correlation coefficients (ICC2,k) and Bland-Altman plots were used to validate the semi-automated technique to the manual segmentation for assessing average cartilage thickness and echo-intensity. A dice correlation coefficient was used to quantify the overlap between the segmentations created with the semi-automated and manual techniques.ResultsFor average cartilage thickness, there was excellent reliability (ICC2,k = 0.99) and a small mean difference (+0.8%) between the manual and semi-automated segmentations. For average echo-intensity, there was excellent reliability (ICC2,k = 0.97) and a small mean difference (−2.5%) between the manual and semi-automated segmentations. The average dice correlation coefficient between the manual segmentation and semi-automated segmentation was 0.90, indicating high overlap between techniques.ConclusionsOur novel semi-automated segmentation technique is a valid method that requires less technical expertise and time than manual segmentation in patients after ACL injury.

Multi-muscle deep learning segmentation to automate the quantification of muscle fat infiltration in cervical spine conditions

Muscle fat infiltration (MFI) has been widely reported across cervical spine disorders. The quantification of MFI requires time-consuming and rater-dependent manual segmentation techniques. A convolutional neural network (CNN) model was trained to segment seven cervical spine muscle groups (left and right muscles segmented separately, 14 muscles total) from Dixon MRI scans (n = 17, 17 scans < 2 weeks post motor vehicle collision (MVC), and 17 scans 12 months post MVC). The CNN MFI measures demonstrated high test reliability and accuracy in an independent testing dataset (n = 18, 9 scans < 2 weeks post MVC, and 9 scans 12 months post MVC). Using the CNN in 84 participants with scans < 2 weeks post MVC (61 females, 23 males, age = 34.2 ± 10.7 years) differences in MFI between the muscle groups and relationships between MFI and sex, age, and body mass index (BMI) were explored. Averaging across all muscles, females had significantly higher MFI than males (p = 0.026). The deep cervical muscles demonstrated significantly greater MFI than the more superficial muscles (p < 0.001), and only MFI within the deep cervical muscles was moderately correlated to age (r > 0.300, p ≤ 0.001). CNN’s allow for the accurate and rapid, quantitative assessment of the composition of the architecturally complex muscles traversing the cervical spine. Acknowledging the wider reports of MFI in cervical spine disorders and the time required to manually segment the individual muscles, this CNN may have diagnostic, prognostic, and predictive value in disorders of the cervical spine.

APPLAUSE: Automatic Prediction of PLAcental health via U-net Segmentation and statistical Evaluation.

Artificial-intelligence population-based automated quantification of placental maturation and health from a rapid functional Magnetic Resonance scan. The placenta plays a crucial role for any successful human pregnancy. Deviations from the normal dynamic maturation throughout gestation are closely linked to major pregnancy complications. Antenatal assessment in-vivo using T2* relaxometry has shown great promise to inform management and possible interventions but clinical translation is hampered by time consuming manual segmentation and analysis techniques based on comparison against normative curves over gestation. This study proposes a fully automatic pipeline to predict the biological age and health of the placenta based on a free-breathing rapid (sub-30 second) T2* scan in two steps: Automatic segmentation using a U-Net and a Gaussian process regression model to characterize placental maturation and health. These are trained and evaluated on 108 3T MRI placental data sets, the evaluation included 20 high-risk pregnancies diagnosed with pre-eclampsia and/or fetal growth restriction. An independent cohort imaged at 1.5T is used to assess the generalization of the training and evaluation pipeline. Across low- and high-risk groups, automatic segmentation performs worse than inter-rater performance (mean Dice coefficients of 0.58 and 0.68, respectively) but is sufficient for estimating placental mean T2* (0.986 Pearson Correlation Coefficient). The placental health prediction achieves an excellent ability to differentiate cases of placental insufficiency between 27 and 33 weeks. High abnormality scores correlate with low birth weight, premature birth and histopathological findings. Retrospective application on a different cohort imaged at 1.5T illustrates the ability for direct clinical translation. The presented automatic pipeline facilitates a fast, robust and reliable prediction of placental maturation. It yields human-interpretable and verifiable intermediate results and quantifies uncertainties on the cohort-level and for individual predictions. The proposed machine-learning pipeline runs in close to real-time and, deployed in clinical settings, has the potential to become a cornerstone of diagnosis and intervention of placental insufficiency. APPLAUSE generalizes to an independent cohort imaged at 1.5T, demonstrating robustness to different operational and clinical environments.

Multi-Muscle Deep Learning Segmentation to Automate the Quantification of Muscle Fat Infiltration in Cervical Spine Conditions

Muscle fat infiltration (MFI) is observed in participants with cervical spine disorders using magnetic resonance imaging. The quantification of MFI using magnetic resonance imaging (MRI) requires time-consuming and rater-dependent manual segmentation techniques. Convolutional neural network (CNN) models have demonstrated state-of-the-art performance in a single cervical spine muscle segmentation task in participants with whiplash injury following a motor vehicle collision (MVC). Here, we expanded this work and trained a CNN for multi-muscle segmentation and automatic MFI calculation of seven cervical spine muscle groups (left and right muscles segmented separately, 14 muscles total) from high-resolution fat-water imaging in participants following a MVC (34 training datasets, n = 17, 14 females, 3 males, age = 33.7 ± 11.4 years). First, we demonstrate high test reliability and accuracy of the CNN MFI measures compared to manual segmentation in an independent testing dataset (n = 18, 11 females, 7 males, age = 31.7 ± 9.6 years). Then in 84 participants imaged within two weeks following a MVC (61 females, 23 males, age = 34.2 ± 10.7 years), we examined MFI across the muscle groups and explored the relationship between MFI and sex, age, and body mass index (BMI). Averaging across all muscle groups, females had significantly higher MFI than males (p 0.300, p

Retinal Vessel Segmentation Using Deep Learning: A Review

This paper presents a comprehensive review of retinal blood vessel segmentation based on deep learning. The geometric characteristics of retinal vessels reflect the health status of patients and help to diagnose some diseases such as diabetes and hypertension. The accurate diagnosis and timing treatment of these diseases can prevent global blindness of patients. Recently, deep learning algorithms have been rapidly applied to retinal vessel segmentation due to their higher efficiency and accuracy, when compared with manual segmentation and other computer-aided diagnosis techniques. In this work, we reviewed recent publications for retinal vessel segmentation based on deep learning. We surveyed these proposed methods especially the network architectures and figured out the trend of models. We summarized obstacles and key aspects for applying deep learning to retinal vessel segmentation and indicated future research directions. This article will help researchers to construct more advanced and robust models.

Comparative analysis of segmentation techniques on MRI brain tumor images

Image Segmentation is considered a necessary technique in medical image processing in extracting the required features for identifying abnormalities in the scanned image. The segmentation of Magnetic Resonance (MR) Brain images helps the radiologist or clinician to classify the tumors or other lesion with accuracy. The manual segmentation technique requires experienced personnel or experts in the classification of brain tumors. To overcome this hindrance many automatic segmentation techniques are introduced. In this proposed work, a comparative study is made on different segmentation techniques like Threshold based, Edge detection, Region growing, Watershed, and K-mean clustering techniques. The mentioned techniques are evaluated and the performance of the different methods are qualitatively analyzed by classifying the tumors from the MR brain images based on the area of segmentation, number of pixels, processing time, and so on.

PWD-3DNet: A Deep Learning-Based Fully-Automated Segmentation of Multiple Structures on Temporal Bone CT Scans

The temporal bone is a part of the lateral skull surface that contains organs responsible for hearing and balance. Mastering surgery of the temporal bone is challenging because of this complex and microscopic three-dimensional anatomy. Segmentation of intra-temporal anatomy based on computed tomography (CT) images is necessary for applications such as surgical training and rehearsal, amongst others. However, temporal bone segmentation is challenging due to the similar intensities and complicated anatomical relationships among critical structures, undetectable small structures on standard clinical CT, and the amount of time required for manual segmentation. This paper describes a single multi-class deep learning-based pipeline as the first fully automated algorithm for segmenting multiple temporal bone structures from CT volumes, including the sigmoid sinus, facial nerve, inner ear, malleus, incus, stapes, internal carotid artery and internal auditory canal. The proposed fully convolutional network, PWD-3DNet, is a patch-wise densely connected (PWD) three-dimensional (3D) network. The accuracy and speed of the proposed algorithm was shown to surpass current manual and semi-automated segmentation techniques. The experimental results yielded significantly high Dice similarity scores and low Hausdorff distances for all temporal bone structures with an average of 86% and 0.755 millimeter (mm), respectively. We illustrated that overlapping in the inference sub-volumes improves the segmentation performance. Moreover, we proposed augmentation layers by using samples with various transformations and image artefacts to increase the robustness of PWD-3DNet against image acquisition protocols, such as smoothing caused by soft tissue scanner settings and larger voxel sizes used for radiation reduction. The proposed algorithm was tested on low-resolution CTs acquired by another center with different scanner parameters than the ones used to create the algorithm and shows potential for application beyond the particular training data used in the study.

Assessment of Racial/Ethnic Disparities in Volumetric MRI Correlates of Clinical Disability in Multiple Sclerosis: A Preliminary Study.

Although global and regional brain volume has been established as a relevant measure to define and predict multiple sclerosis (MS) severity, characterization of specific trends by race/ethnicity is currently lacking. We aim to (1) characterize racial disparities in disability-specific patterns of brain MRI volumetric measures between Hispanic and Caucasian individuals with MS and (2) explore the relevance of these measures as predictors of clinical disability progression. Brain MRI scans from 94 Hispanic and 94 age- and gender-matched Caucasian MS patients were analyzed using automatic and manual segmentation techniques. Select global and regional volume measures were correlated to Expanded Disability Status Scale (EDSS) scores at baseline and subsequent follow-up visits. Hispanic patients had a higher baseline median EDSS score (interquartile range [IQR], 2.0; [1.0-3.5]) compared to Caucasians (median [IQR], 1.0 [.0-2.0]) and an increased risk of requiring ambulatory assistance (hazard ratio [HR], 9.7; 95% confidence interval [CI], 2.8-32.5). Normalized thalamic volume was moderately associated with EDSS scores (rs = -.42, P < .001 in Hispanics;rs = -.32,P = .002 in Caucasians) and was the best predictor of sustained disability worsening in both racial groups in a time-to-event analysis. The confounding impact of race on quantitative brain volume measures may affect the interpretation of outcome measures in MS clinical trials.

Application of 3-dimensional reflectance confocal microscopy: Melanocytic proliferations as three-dimensional models

To the Editor: Melanoma on chronically sun-damaged skin (MSDS) is challenging to detect.1Sanchez M.I. Rabinovitz H.S. Oliverio M.C. Elgart G.W. Grichnik J.M. Confocal microscopy of challenging dermoscopic diagnoses.Arch Dermatol. 2012; 148: 1224Crossref PubMed Scopus (1) Google Scholar Reflectance confocal microscopy (RCM) increases diagnostic accuracy but relies upon interpretation of 2-dimensional (2D) horizontal sections. Our objective was to determine whether existing technology could be used to create 3-dimensional (3D) models, which would expand on data available for discriminating skin biology. Melanocytic cells have different morphologic appearances dependent on developmental stage, activation status, and pathogenic genetic and epigenetic alterations. Cell shape, therefore, can provide clues to a cell's behavior. We reconstructed individual melanoma cells from stacks of RCM images (Fig 1; Supplemental Fig 1, available via Mendeley at https://data.mendeley.com/datasets/7kx38d9h2v/3). Distinct morphologies could be identified, including spindle, spheroid, and heavily dendritic cell forms. These 3D reconstructions also allowed for modeling of the nuclear shapes (Fig 1, insets). Because image sizes are known, it is possible to calculate nuclear and cellular volumes, which may also increase diagnostic accuracy.2Chang C.K. Tsai C.C. Hsu W.Y. et al.Segmentation of nucleus and cytoplasm of a single cell in three-dimensional tomogram using optical coherence tomography [published correction appears in J Biomed Opt. 2017;22(3):39801].J Biomed Opt. 2017; 22: 36003Crossref PubMed Scopus (11) Google Scholar Monomorphism or pleomorphism of melanocytes may also provide clues to diagnosis. In some tumors, cells are relatively uniform, presenting as sheets of spindle-shaped cells, monomorphic spheroidal-shaped cells, or heavy mats of dendrites. However, other tumors display more variation with epithelioid, spindle-shaped, and dendritic morphologies all appearing in a 500-μm field (Supplemental Fig 2). Location is also important in diagnosis. The presence of melanocytic cells in the middle to upper epidermis (pagetoid spread) may suggest malignancy, whereas cells in the superficial dermis suggest invasion. Within the context of the 3D volume, we used renderings to show upper (epidermal) and lower (dermal) regions simultaneously (Supplemental Fig 3). This can be used not only for identification of pagetoid and invasive cellular locations but also for examining the extent of melanization of cells in different locations that may imply differences in maturation. When visualizing normal pigmented skin and some benign pigmented neoplasms with RCM, bright rims (edging) around the dermal papillae are visible as rings when viewed in the horizontal plane, primarily due to melanin-containing basal keratinocytes. By rendering the stack on the vertical axis, the normal arrangement of refractile cells along the papillae is apparent (Fig 2, Supplemental Fig 4). Epidermal thickness can be calculated, as well as the length of epidermal ridge and volume of dermal papillae in 3D. When MSDS is present, asymmetric pigmentation (partially edged or nonedged dermal papillae) may be noted around dermal papillae.3Serrano P.F.-C. Segurado-Miravalles G. González S. Reflectance confocal microscopy in oncological dermatology.in: Nouri K. Lasers in Dermatology and Medicine. Springer, 2018: 375-400Google Scholar This phenomenon may be due to nonuniform melanin transfer or melanoma cell population density differences. The dermoepidermal junction also frequently becomes flattened in malignant tumors. These phenomena may be best evaluated in 3D. The 3D modeling process currently has limitations. When manual segmentation techniques are used, the more complex models take hours to produce, limiting bedside use. The imaging area is limited to the stack dimensions (500 μm × 500 μm). RCM image stacks are anisotropic, thus the noncubic voxels require interpolation and resampling to render accurately in 3D space.4Trusk T.C. 3D reconstruction of confocal image data.in: Gray W. Price J.R.L. Basic Confocal Microscopy. Springer, Cham2018: 279-307Crossref Google Scholar In summary, 3D modeling of confocal images can provide supplementary information, beyond what is appreciable in 2D images, including melanocyte morphology, volume, pleomorphism, and location. 3D modeling also allows for detailed evaluation of the dermoepidermal contour, which may reflect the summation of keratinocytic, melanocytic, immunologic, and microenvironmental events.