Manual 3D Research Articles

CT-based radiomics model with machine learning for predicting primary treatment failure in diffuse large B-cell Lymphoma

Biomarkers which can identify Diffuse Large B-Cell Lymphoma (DLBCL) likely to be refractory to first-line therapy are essential for selecting this population prior to therapy initiation to offer alternate therapeutic options that can improve prognosis. We tested the ability of a CT-based radiomics approach with machine learning to predict Primary Treatment Failure (PTF)-DLBCL from initial imaging evaluation. Twenty-six refractory patients were matched to 26 non-refractory patients, yielding 180 lymph nodes for analysis. Manual 3D delineation of the total node volume was performed by two independent readers to test the reproducibility. Then, 1218 hand-crafted radiomic features were extracted. The Random Forests machine learning approach was used as a classifier for constructing the prediction models. Seventy percent of the nodes were randomly assigned to a training set and the remaining 30% were assigned to an independent test set. The final model was tested on the dataset from the 2 readers, showing a mean accuracy, sensitivity and specificity of 73%, 62% and 82%, respectively, for distinguishing between refractory and non-refractory patients. The area under the receiver operating characteristic curve (AUC) was 0.83 and 0.79 for the two readers. We conclude that machine learning CT-based radiomics analysis is able to identify a priori PTF-DLBCL with a good accuracy.

Optimal Woven EndoBridge (WEB) Device Size Selection Using Automated Volumetric Software.

Introduction: Selecting the appropriate Woven EndoBridge (WEB) device sizing for the treatment of wide-neck bifurcation aneurysms (WNBAs) remains challenging. The aim of this study was to evaluate different volumetric-based imaging methodologies to predict an accurate WEB device size selection to result in a successful implantation. Methods: All consecutive patients treated with WEB devices for intracranial aneurysms from January 2019 to June 2020 were included. Aneurysm dimensions to calculate aneurysm volumes were measured using three different modalities: automated three-dimensional (3D) digital subtraction angiography (DSA), manual 3D DSA, and two-dimensional (2D) DSA. The device–aneurysm volume (DAV) ratio was defined as device volume divided by the aneurysm volume. WEB volumes and the DAV ratios were used for assessing the device implantation success and follow-up angiographic outcomes at six months. Pearson correlation, Wilcoxon Rank Sum test, and density approximations were used for estimating the WEB volumes and the imaging modality volumes for successful implantation. Results: A total of 41 patients with 43 aneurysms were included in the study. WEB device and aneurysm volume correlation coefficient was highest for 3D automatic (r = 0.943), followed by 3D manual (r = 0.919), and 2D DSA (r = 0.882) measurements. Measured median volumes were significantly different for 3D automatic and 2D DSA (p = 0.017). The highest rate of successful implantation (87.5%) was between 0.6 and 0.8 DAV ratio. Conclusion: Pre-procedural assessment of DAV ratios may increase WEB device implantation success. Our results suggest that volumetric measurements, especially using automated 3D volumes of the aneurysms, can assist in accurate WEB device size selection.

PHOTOGRAMMETRY AND COMPUTED TOMOGRAPHY POINT CLOUD REGISTRATION USING VIRTUAL CONTROL POINTS

Abstract. In this paper we propose a virtual control point based method for the registration of photogrammetry and computed tomography (CT) data. Because of the fundamentally different two data sources, conventional registration methods, such as manual control points registration or 3D local feature-based registration, are not suitable. The registration objective of our application is about 3D reconstructions of gyroscopes, which contain abundant geometric primitives to be fitted in the point clouds. In the first place, photogrammetry and CT scanning are applied, respectively, for 3D reconstructions. Secondly, our workflow implements a segmentation after obtaining the surface point cloud from the complete CT volumetric data. Then geometric primitives are fitted in this point cloud benefitting from the less complex cluster segments. In the next step, intersection operations of the parametrized primitives generates virtual points, which are utilized as control points for the transformation parameters estimation. A random sample consensus (RANSAC) method is applied to find the correspondences of both virtual control point sets using corresponding descriptors and calculates the transformation matrix as an initial alignment for further refining the registration. The workflow is invariant to pose, resolution, completeness and noise within our validation process.

Balance, barycentremetry and external shape analysis in idiopathic scoliosis: What can the physician expect from it?

ObjectiveOur objective was to establish a corridor of normality for the external shape 3D parameters and then to assess these variables in adolescent idiopathic scoliosis (AIS). MethodsAdolescent with mild and severe AIS were included prospectively, as well as a control group of asymptomatic subjects. A quasi‐automatic 3D reconstruction of the spine and manual 3D reconstruction of the external envelope was performed from biplanar radiography. The center of mass position, the axial intersegmental moment resulting at the apex and junctional vertebrae, and the coronal trunk balance were automatically computed. A normality corridor of asymptomatic subjects was calculated as the range [5th-95th percentiles] for external shape parameters at each vertebral level. ResultsForty-one asymptomatic subjects (19 females; 22 males; 21 yo, SD=4) and sixty AIS (56 females; 4 males; 13 years old, SD=1.9; 30 mild and 30 severe; 34 thoracic curves and 26 thoraco-lumbar or lumbar curves) were included. All parameters based on the external shape showed differences between AIS and controls, as well as between mild and severe scoliosis. For instance, the intersegmental moment applied to the upper junctional vertebra was above the 95th percentile of controls in 70% of AIS patient. The percentage of severe patients showing parameters higher than the normality corridor was significantly higher than mild patients (p<0.0001). ConclusionsThe analysis of center of mass, vertebral intersegmental moment and coronal trunk balance parameters appear to be relevant in characterizing the 3D deformity of adolescent idiopathic scoliosis. The upper junctional intersegmental moment seems to be able to distinguish the different stages of curvature severity.

A data-driven approach towards the full anthropometric measurements prediction via Generalized Regression Neural Networks

Anthropometry is a science concerning human body dimension measurements and proportions that is widely applied in multiple disciplines, particularly in apparel and “human-centered” product design. Designers and engineers generally use percentile anthropometric data that leads to significant errors in practice. Although adopting non-statistic body dimensions could resolve this issue, gaining detailed body measurements is a challenging task as neither manual measurement nor 3D scanning is efficient in a cost-effective manner. With the rapid development of artificial neural networks, predicting body sizes and shapes, instead of measuring them, has become a new trend. This work presents a unique Generalized Regression Neural Network architecture that is capable of accurately predicting 76 detailed body measurements from seven easily measured body features with high tolerance to input measurement errors. The proposed model outperforms the existing regression models and can be easily implemented in the design process.

Developing and Validating a Computer-Based Training Tool for Inferring 2D Cross-Sections of Complex 3D Structures.

Developing and validating a novel domain-agnostic, computer-based training tool for enhancing 2D cross-section understanding of complex 3D structures. Understanding 2D cross-sections of 3D structures is a crucial skill in many disciplines, from geology to medical imaging . It requires a complex set of spatial/visualization skills including mental rotation, spatial structure understanding, and viewpoint projection. Prior studies show that experts differ from novices in these skills. We have developed a novel training tool for inferring 2D cross-sections of 3D structures using a participatory design methodology. We used a between-subject study design, with 60 participants, to evaluate the training tool. Our primary effectiveness evaluation was based on pre- and postspatial tests that measured both cross-section abilities and specific spatial skills: viewpoint, mental rotation, and card rotation. Results showed significant performance gains on inferring 2D cross-sections for participants of the training group. Our tool improves two other spatial skills as well: mental rotation and viewpoint visualization. Our training tool was effective not only in enhancing 2D cross-section understanding of complex 3D structures, but also in improving mental rotation and viewpoint visualization skills. Our tool can be beneficial in different fields such as medical imaging, biology, geology, and engineering. For example, an application of our tool is in medical/research labs to train novice segmenters in ongoing manual 3D segmentation tasks. It can also be adapted in other contexts, such as training children, older adults, and individuals with very low spatial skills.

Automatic segmentation of left and right ventricles in cardiac MRI using 3D-ASM and deep learning

Segmentation of the left and right ventricles in cardiac MRI (Magnetic Resonance Imaging) is a prerequisite step for evaluating global and regional cardiac function. This work presents a novel and robust schema for MRI segmentation by combining the advantages of deep learning localization and 3D-ASM (3D Active Shape Model) restriction without any user interaction. Three fundamental techniques are exploited: (1) manual 2D contours are used to build distance maps to get 3D ground truth shape, (2) derived right ventricle points are employed to rotate the coarse initial shape for a refined bi-ventricle initial estimation, (3) segmentation results from deep learning are utilised to build distance maps for the 3D-ASM matching process to help image intensity modelling. The datasets used for experimenting the cine MRI data are 1000 cases from UK Biobank, 500 subjects are selected to train CNN (Convolution Neural Network) parameters, and the remaining 500 cases are adopted for validation. Specifically, cases are used to rebuild point distribution and image intensity models, and also utilized to train CNN. In addition, the left 500 cases are used to perform the validation experiments. For the segmentation of the RV (Right Ventricle) endocardial contour, LV (Left Ventricle) endo- and epicardial contours, overlap, Jaccard similarity index, Point-to-surface errors and cardiac functional parameters are calculated. Experimental results show that the proposed method has advantages over the previous approaches.

Scratch Teaching Mode of a Course for College Students

Maker education is of great significance for the fostering of innovative talents. At present, however, most of the maker courses in colleges and universities are with such problems as single teaching functions, non-uniform course management system and unsmooth course teaching. Therefore, how to innovate in concept, system and mode to adapt them to the new demand of the development of higher education is a realistic problem to be addressed by maker course. Guided by the teaching concept of STEAM, our study establishes maker education oriented on learners and implements the hands-on operating experience of “creating, delighting, cooperating and sharing”. Meanwhile, the teaching design of Scratch programming education dominated by game-based learning is taken as an auxiliary teaching measure. Moreover, four teaching modules, i.e., design development teaching, artistic design teaching, manual production teaching and 3D printing teaching, are also created. On this basis, a model for the development elements of the thinking of maker education is proposed to evaluate the training effect of innovation ability by this teaching practice. The results show that the proposed teaching mode can effectively improve students’ learning ability and innovative thinking. Hopefully, the present study can offer a reference for researchers and practitioners of maker education.

A detailed review of forensic facial reconstruction techniques.

Facial reconstruction is an effective forensic technique that can help recreate a victim's facial appearance from the skull. It is typically used to assist law enforcement agencies to identify missing deceased persons. Reconstruction techniques are usually based on the relationship between the underlying hard tissues, such as bone structure, and soft tissues such as the facial muscles and facial features. Facial reconstruction can be a feasible alternative to identify the remains from a decomposed, mutilated, or skeletonised corpse. It is important to remember that although the outcomes are empirical in nature, the technique has been applied widely in many situations. Recent advancements in technology and computer-based techniques have increased the accuracy and validity of this forensic discipline. We consider the most commonly used facial reconstruction techniques in this paper, with a detailed description of manual 3D techniques.

Handcrafted MRI radiomics and machine learning: Classification of indeterminate solid adrenal lesions

PurposeTo assess a radiomic machine learning (ML) model in classifying solid adrenal lesions (ALs) without fat signal drop on chemical shift (CS) as benign or malignant. Method55 indeterminate ALs (21 lipid poor adenomas, 15 benign pheocromocytomas, 1 oncocytoma, 12 metastases, 6 primary tumors) showing no fat signal drop on CS were retrospectively included. Manual 3D segmentation on T2-weighted and CS images was performed for subsequent radiomic feature extraction. After feature stability testing and an 80–20% train-test split, the train set was balanced via oversampling. Following a multi-step feature selection, an Extra Trees model was tuned with 5-fold stratified cross-validation in the train set and then tested on the hold-out test set. ResultsA total of 3396 features were extracted from each AL, of which 133 resulted unstable while none had low variance (< 0.01). Highly correlated (r > 0.8) features were also excluded, leaving 440 parameters. Among these, Support Vector Machine 5-fold stratified cross-validated recursive feature elimination selected a subset of 6 features. ML obtained a cross-validation accuracy of 0.94 on the train and 0.91 on the test sets. Precision, recall and F1 score were respectively 0.92, 0.91 and 0.91. ConclusionsOur MRI handcrafted radiomics and ML pipeline proved useful to characterize benign and malignant solid indeterminate adrenal lesions.

Abstract PO-017: Annotation-free 3D gland segmentation with generative image-sequence translation for prostate cancer risk assessment

Abstract 1. Purpose We developed an annotation-free image-translation-based approach for 3D segmentation of prostate glands within 3D-histology datasets of whole biopsies stained with a low-cost and rapid-diffusing fluorescent analog of H&amp;E. 2. Introduction The current diagnostic gold standard of histopathology provides a limited 2D view of complex 3D glandular structures in prostate specimens, which contributes to high interobserver variability and reduced prognostic accuracy. We have recently developed nondestructive 3D pathology methods based on open-top light-sheet (OTLS) microscopy. In order to train prognostic models based on 3D glandular morphology, we have developed an objective (biomarker-based) 3D gland-segmentation method that does not rely upon tedious and subjective manual annotations, and which can operate on images of tissue stained with a cheap and fast-diffusing (small molecule) fluorescent analog of H&amp;E. 3. Methods We first convert H&amp;E-analog images into synthetic 3D immunofluorescence (IF) images of cytokeratin-8 (CK8), a biomarker expressed by the luminal epithelial cells that surround all prostate glands. This conversion was performed by treating the 3D data as sequences of 2D images and adapting a generative adversarial network (GAN)-based video-synthesis model to perform image-sequence translation with high depth-wise continuity between frames. Based on the synthetic CK8 images, glands were then objectively segmented in 3D using a thresholding/morphology-based algorithm. This two-step method obviates the need for labor-intensive and subjective manual 3D annotations, as would be needed to train a single-step segmentation model. A 3D structural similarity (SSIM) index was assessed between synthetic and real CK8 images. In addition, based on ground-truth manual annotations of glands in ten 0.2-mm3 regions, we calculated voxel-based Dice coefficients to compare our segmentation accuracy vs. two baseline methods. 4. Results The synthetic CK8 images exhibited high fidelity (3D SSIM = 0.696) and optimal continuity with depth. Our segmentation accuracy outperformed two baseline methods, with Dice coefficients (averaged for 10 samples) of: 0.882 (our method), 0.725 (3D watershed), 0.643 (2D U-net). We are now applying our method to 3D histology datasets of whole biopsies (n &amp;gt; 1000) acquired ex vivo from prostatectomy specimens (N ~ 200), where we are extracting 3D histomorphometric gland features to predict biochemical recurrence (BCR) post-prostatectomy in men with prostate cancer. 5. Conclusions Our annotation-free segmentation method relies upon generative synthetic 3D IF images from H&amp;E-analog images in order to objectively segment the 3D prostate gland network. These accurate 3D segmentations are being extended to whole-biopsy 3D pathology datasets for prostate cancer risk assessment. Citation Format: Weisi Xie, Adam Glaser, Nicholas Reder, Nadia Postupna, Chenyi Mao, Can Koyuncu, Patrick Leo, Robert Serafin, Hongyi Huang, Anant Madabhushi, Lawrence True, Jonathan T.C. Liu. Annotation-free 3D gland segmentation with generative image-sequence translation for prostate cancer risk assessment [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-017.

Automatic segmentation and location learning of neonatal cerebral ventricles in 3D ultrasound data combining CNN and CPPN

Preterm neonates are highly likely to suffer from ventriculomegaly, a dilation of the Cerebral Ventricular System (CVS). This condition can develop into life-threatening hydrocephalus and is correlated with future neuro-developmental impairments. Consequently, it must be detected and monitored by physicians. In clinical routing, manual 2D measurements are performed on 2D ultrasound (US) images to estimate the CVS volume but this practice is imprecise due to the unavailability of 3D information. A way to tackle this problem would be to develop automatic CVS segmentation algorithms for 3D US data. In this paper, we investigate the potential of 2D and 3D Convolutional Neural Networks (CNN) to solve this complex task and propose to use Compositional Pattern Producing Network (CPPN) to enable Fully Convolutional Networks (FCN) to learn CVS location. Our database was composed of 25 3D US volumes collected on 21 preterm nenonates at the age of 35.8±1.6 gestational weeks. We found that the CPPN enables to encode CVS location, which increases the accuracy of the CNNs when they have few layers. Accuracy of the 2D and 3D FCNs reached intraobserver variability (IOV) in the case of dilated ventricles with Dice of 0.893±0.008 and 0.886±0.004 respectively (IOV = 0.898±0.008) and with volume errors of 0.45±0.42 cm3 and 0.36±0.24 cm3 respectively (IOV = 0.41±0.05 cm3). 3D FCNs were more accurate than 2D FCNs in the case of normal ventricles with Dice of 0.797±0.041 against 0.776±0.038 (IOV = 0.816±0.009) and volume errors of 0.35±0.29 cm3 against 0.35±0.24 cm3 (IOV = 0.2±0.11 cm3). The best segmentation time of volumes of size 320×320×320 was obtained by a 2D FCN in 3.5±0.2 s.

Deep Learning Corpus Callosum Segmentation as a Neurodegenerative Marker in Multiple Sclerosis.

Corpus callosum atrophy is a sensitive biomarker of multiple sclerosis (MS) neurodegeneration but typically requires manual 2D or volumetric 3D-based segmentations. We developed a supervised machine learning algorithm, DeepnCCA, for corpus callosum segmentation and relate callosal morphology to clinical disability using conventional MRI scans collected in clinical routine. In a prospective study of 553 MS patients with 704 acquisitions, 200 unique 2D T2 -weighted MRI scans were delineated to develop, train, and validate DeepnCCA. Comparative FreeSurfer segmentations were obtained in 504 3D T1 -weighted scans. Both FreeSurfer and DeepnCCA outputs were correlated with clinical disability. Using principal component analysis of the DeepnCCA output, the morphological changes were explored in relation to clinical disease burden. DeepnCCA and manual segmentations had high similarity (Dice coefficients 98.1 .11%, 89.3 .76%, for intracranial and corpus callosum area, respectively through 10-fold cross-validation). DeepnCCA had numerically stronger correlations with cognitive and physical disability as compared to FreeSurfer: Expanded disability status scale (EDSS) ±6 months (r = -.22 P = .002; r = -.17, P = .013), future EDSS (r = -.26, P<.001; r = -.17, P = .012), and future symbol digit modalities test (r = .26, P = .001; r = .24, P = .003). The corpus callosum became thinner with increasing cognitive and physical disability. Increasing physical disability, additionally, significantly correlated with a more angled corpus callosum. DeepnCCA (https://github.com/plattenmichael/DeepnCCA/) is an openly available tool that can provide fast and accurate corpus callosum measurements applicable to large MS cohorts, potentially suitable for monitoring disease progression and therapy response.

Complementary Phenotyping of Maize Root System Architecture by Root Pulling Force and X-Ray Imaging

The root system is critical for the survival of nearly all land plants and a key target for improving abiotic stress tolerance, nutrient accumulation, and yield in crop species. Although many methods of root phenotyping exist, within field studies, one of the most popular methods is the extraction and measurement of the upper portion of the root system, known as the root crown, followed by trait quantification based on manual measurements or 2D imaging. However, 2D techniques are inherently limited by the information available from single points of view. Here, we used X-ray computed tomography to generate highly accurate 3D models of maize root crowns and created computational pipelines capable of measuring 71 features from each sample. This approach improves estimates of the genetic contribution to root system architecture and is refined enough to detect various changes in global root system architecture over developmental time as well as more subtle changes in root distributions as a result of environmental differences. We demonstrate that root pulling force, a high-throughput method of root extraction that provides an estimate of root mass, is associated with multiple 3D traits from our pipeline. Our combined methodology can therefore be used to calibrate and interpret root pulling force measurements across a range of experimental contexts or scaled up as a stand-alone approach in large genetic studies of root system architecture.

Automated Follicular Assessment Using a Novel Two-dimensional Ultrasound-based Solution.

Background:High intra- and interobserver variability in the follicular assessment using two-dimensional (2D) ultrasound (US) is still a concern. To solve this issue, we have developed a novel software solution, which automatically provides follicles' count and their diameters using 2D US images obtained by a manual sweep of an ovary. The primary objective of this study was to compare the result of the automated solution with a manual 2D US-based assessment.Methods:In the first phase, multiple follicular US sweeps were collected from 54 subjects; these sweeps were used to develop the software. In the second phase, data from 10 subjects were collected for validation of the developed solution. During each phase, for follicles ≥5 mm, their count and diameters were recorded by the sonologist using 2D US.Results:For the total follicle count, a high correlation (0.787) was observed between the solution and manual assessment. The 95% limits of agreement between the two methods were in the range of 4.232 to −4.258. The two methods had an excellent correlation (0.817) for the measurement of mean follicular diameter. However, the solution had a tendency to underestimate the mean diameter by an average of 1.725 mm (±2.16 mm). The limits of agreement between the two methods for mean diameter measurement were from 2.508 to −5.960 mm.Conclusion:This study validates the feasibility of our solution for automatic assessment of follicle count and diameter with accuracy comparable to the 2D US-based manual assessment. We further observed that the solution's performance is better than known intra- and interobserver variability of the manual assessment. We recommend further validation of the solution to confirm these initial results and potential time gain with an automated assessment.

Learning From Multiple Datasets With Heterogeneous and Partial Labels for Universal Lesion Detection in CT.

Large-scale datasets with high-quality labels are desired for training accurate deep learning models. However, due to the annotation cost, datasets in medical imaging are often either partially-labeled or small. For example, DeepLesion is such a large-scale CT image dataset with lesions of various types, but it also has many unlabeled lesions (missing annotations). When training a lesion detector on a partially-labeled dataset, the missing annotations will generate incorrect negative signals and degrade the performance. Besides DeepLesion, there are several small single-type datasets, such as LUNA for lung nodules and LiTS for liver tumors. These datasets have heterogeneous label scopes, i.e., different lesion types are labeled in different datasets with other types ignored. In this work, we aim to develop a universal lesion detection algorithm to detect a variety of lesions. The problem of heterogeneous and partial labels is tackled. First, we build a simple yet effective lesion detection framework named Lesion ENSemble (LENS). LENS can efficiently learn from multiple heterogeneous lesion datasets in a multi-task fashion and leverage their synergy by proposal fusion. Next, we propose strategies to mine missing annotations from partially-labeled datasets by exploiting clinical prior knowledge and cross-dataset knowledge transfer. Finally, we train our framework on four public lesion datasets and evaluate it on 800 manually-labeled sub-volumes in DeepLesion. Our method brings a relative improvement of 49% compared to the current state-of-the-art approach in the metric of average sensitivity. We have publicly released our manual 3D annotations of DeepLesion online.1 1https://github.com/viggin/DeepLesion_manual_test_set.

MRI-based 3D models of the hip joint enables radiation-free computer-assisted planning of periacetabular osteotomy for treatment of hip dysplasia using deep learning for automatic segmentation

IntroductionBoth Hip Dysplasia(DDH) and Femoro-acetabular-Impingement(FAI) are complex three-dimensional hip pathologies causing hip pain and osteoarthritis in young patients. 3D-MRI-based models were used for radiation-free computer-assisted surgical planning. Automatic segmentation of MRI-based 3D-models are preferred because manual segmentation is time-consuming.To investigate(1) the difference and(2) the correlation for femoral head coverage(FHC) between automatic MR-based and manual CT-based 3D-models and (3) feasibility of preoperative planning in symptomatic patients with hip diseases. MethodsWe performed an IRB-approved comparative, retrospective study of 31 hips(26 symptomatic patients with hip dysplasia or FAI). 3D MRI sequences and CT scans of the hip were acquired. Preoperative MRI included axial-oblique T1 VIBE sequence(0.8 mm3 isovoxel) of the hip joint. Manual segmentation of MRI and CT scans were performed. Automatic segmentation of MRI-based 3D-models was performed using deep learning. Results(1)The difference between automatic and manual segmentation of MRI-based 3D hip joint models was below 1 mm(proximal femur 0.2 ± 0.1 mm and acetabulum 0.3 ± 0.5 mm). Dice coefficients of the proximal femur and the acetabulum were 98 % and 97 %, respectively. (2)The correlation for total FHC was excellent and significant(r = 0.975, p < 0.001) between automatic MRI-based and manual CT-based 3D-models. Correlation for total FHC (r = 0.979, p < 0.001) between automatic and manual MR-based 3D models was excellent.(3)Preoperative planning and simulation of periacetabular osteotomy was feasible in all patients(100 %) with hip dysplasia or acetabular retroversion. ConclusionsAutomatic segmentation of MRI-based 3D-models using deep learning is as accurate as CT-based 3D-models for patients with hip diseases of childbearing age. This allows radiation-free and patient-specific preoperative simulation and surgical planning of periacetabular osteotomy for patients with DDH.

Evaluation of a Weightbearing CT Artificial Intelligence-based Automatic Measurement for Hallux Valgus: A Case-Control Study

Category:Bunion; OtherIntroduction/Purpose:Cone Beam Weight Bearing CT (WBCT) is gaining traction, particularly in the foot and ankle, due to the ability to perform natural stance weight bearing 3D scans. However, the resulting wealth of 3D data renders daily clinical use time consuming. Therefore, reliable automatic measurements are indispensable in order to make best use of the technology. The aim of this study was to evaluate a beta-version WBCT artificial intelligence (AI) automatic measurement system for the M1-M2 intermetatarsal angle (IMA), which is applicable in the absence of metallic hardware in the foot and ankle. We hypothesized that automatic measurements would correlate well with human measurements, and that software reproducibility would be better and close to perfect compared to manual measurements.Methods:In this retrospective case-control study, 90 feet were included from patients who underwent WBCT scans during routine follow up: 44 feet (90.9% female, mean age 54 years) with symptomatic hallux valgus (HV), 46 controls (76.1% female, mean age 49 years). Patients were excluded if they had history of surgery or trauma involving the first or second metatarsals, hallux rigidus, or presence of metal in their foot/ankle. IMA was measured manually on Digitally Reconstructed Radiographs (DRR IMA) and automatically with AI software producing auto 2D (ground plane projection) and 3D (multiplanar) measurements. Each IMA DRR was measured by two independent raters twice to calculate intraclass correlation coefficients (ICCs). To assess intra- software reliability, AI software measurements were made twice on each dataset. Manual and automatic measurements were compared between HV and control groups. Failures of the AI software to produce a measurement were recorded.Results:Mean values for controls were 8.6° +-1.8° (range, 5°-14°) for the manually measured DRR IMA, 9.3° +-2.8° (range, 3°- 17°) for auto 2D, and 9.2° +-2.6° (range, 3°-16°) for auto 3D IMA measurements. Compared to controls, HV patients demonstrated significantly increased IMA (p<0.0001): 14.2° +-2.7° (range, 8°-21°) for the manually measured DRR IMA, 15.4°+- 4.4° (range, 8°-26°) for auto 2D, and 15.1° +-4.1° (range 8°-28°) for auto 3D IMA measurements. There were strong correlations (r=0.75 and r=0.80) between manual and auto 2D and 3D measurements. Intraobserver and interobserver ICCs for DRR IMA were 0.96 and 0.90, respectively, and the intra-software ICCs for the AI were near 1.0 for both auto 2D and auto 3D IMA. The AI software failed in 32.3% cases.Conclusion:Our results demonstrated strong correlation between a WBCT Artificial Intelligence based automatic measurement for IMA with human measurements, with the ability to distinguish HV from control with close to 100% repeatability. However, the number of failures was still high due to the early stage beta-version of the algorithm tested. While these early results are promising, further developments are warranted in order to improve usability of this tool in daily practice, especially in the presence of metal hardware.Figure 1.

Navigating the mesentery: Part III. Unusual anatomy of ileocolic vessels.

The ileocolic vessels are important landmarks in advanced surgery of the midgut. The aim of the present study is to present variations of ileocolic vessels relevant to complete mesocolic excision with D3 lymphadenectomy of the right colon, within their detailed and precise morphometric framework and deriving from a large and consistent series of operated patients. An ongoing prospective trial 'Safe Radical D3 Right Hemicolectomy for Cancer through Preoperative Biphasic Multidetector Computed Tomography Angiography' was reviewed. The imaging datasets underwent manual segmentation and 3D reconstruction, and the results were validated at surgery. A total of 356 patients were included in the study. A cross-section of the series revealed five cases (1.4%) with variation of ileocolic vessels relevant to complete mesocolic excision or D3 extended mesenterectomy. There were two cases with absence of a true classical ileocolic artery, two cases with absence of a true classical ileocolic vein, and one case of precocious bifurcation of the ileocolic artery, left to the superior mesenteric vein. The entire D3 area in all the cases was thoroughly documented and analysed from the morphometric point of view (calibres, lengths of vessels, crossing patterns). The preoperative visualization of a patient's individual 3D anatomy is a powerful tool in identifying the variations whose negligence could have dire consequences in complete mesocolic excision of the right colon.

Automated brainstem volumetry can aid in the diagnostics of parkinsonian disorders

IntroductionSeparating progressive supranuclear palsy (PSP) from Parkinson's disease (PD) and multiple system atrophy (MSA) is often challenging in early disease but is important for appropriate management. Magnetic resonance imaging (MRI) can aid the diagnostics and manual 2D measurements are often used. However, new fully automatic brainstem volumetry could potentially be more accurate and increase availability of brainstem metrics. MethodsClinical 3D T1-weighted MRI were obtained from 196 consecutive patients; 29 PSP, 27 MSA, 140 PD. Midbrain-pons ratio and magnetic resonance parkinsonism index (MRPI) 1.0 and 2.0 were manually calculated, and intra-rater and inter-rater reliability was assessed. FreeSurfer was used to automatically segment brainstem substructures, normalized to the intracranial volume. The robustness of the automated analysis was evaluated in 3 healthy controls. The diagnostic accuracy of the brainstem biomarkers was assessed using receiver operating characteristic curves. ResultsAutomatic brainstem volumetry had good repeatability/reproducibility with intra-scanner coefficient of variation 0.3–5.5% and inter-scanner coefficient of variation 0.9–8.4% in the different brainstem regions. Midbrain volume performs better than planimetric measurements in separating PSP from PD (Area under the curve (AUC) 0.90 compared with 0.81 for midbrain-pons ratio (p = 0.019), 0.77 for MRPI 1.0 (p = 0.007) and 0.81 for MRPI 2.0 (p = 0.021)). Midbrain volume performed on par with planimetry for separation between PSP and MSA. ConclusionAutomatic brainstem segmentation is robust and shows promising diagnostic performance in separating PSP from PD and MSA. If further developed, it could play a role in diagnosing PSP and could potentially be used as an outcome in clinical trials.