Average Volume Error Research Articles

Automatic segmentation of lower limb muscles from MR images of post-menopausal women based on deep learning and data augmentation.

Individual muscle segmentation is the process of partitioning medical images into regions representing each muscle. It can be used to isolate spatially structured quantitative muscle characteristics, such as volume, geometry, and the level of fat infiltration. These features are pivotal to measuring the state of muscle functional health and in tracking the response of the body to musculoskeletal and neuromusculoskeletal disorders. The gold standard approach to perform muscle segmentation requires manual processing of large numbers of images and is associated with significant operator repeatability issues and high time requirements. Deep learning-based techniques have been recently suggested to be capable of automating the process, which would catalyse research into the effects of musculoskeletal disorders on the muscular system. In this study, three convolutional neural networks were explored in their capacity to automatically segment twenty-three lower limb muscles from the hips, thigh, and calves from magnetic resonance images. The three neural networks (UNet, Attention UNet, and a novel Spatial Channel UNet) were trained independently with augmented images to segment 6 subjects and were able to segment the muscles with an average Relative Volume Error (RVE) between -8.6% and 2.9%, average Dice Similarity Coefficient (DSC) between 0.70 and 0.84, and average Hausdorff Distance (HD) between 12.2 and 46.5 mm, with performance dependent on both the subject and the network used. The trained convolutional neural networks designed, and data used in this study are openly available for use, either through re-training for other medical images, or application to automatically segment new T1-weighted lower limb magnetic resonance images captured with similar acquisition parameters.

Prediction of tissue outcome in acute ischemic stroke based on single-phase CT angiography at admission.

In acute ischemic stroke, prediction of the tissue outcome after reperfusion can be used to identify patients that might benefit from mechanical thrombectomy (MT). The aim of this work was to develop a deep learning model that can predict the follow-up infarct location and extent exclusively based on acute single-phase computed tomography angiography (CTA) datasets. In comparison to CT perfusion (CTP), CTA imaging is more widely available, less prone to artifacts, and the established standard of care in acute stroke imaging protocols. Furthermore, recent RCTs have shown that also patients with large established infarctions benefit from MT, which might not have been selected for MT based on CTP core/penumbra mismatch analysis. All patients with acute large vessel occlusion of the anterior circulation treated at our institution between 12/2015 and 12/2020 were screened (N = 404) and 238 patients undergoing MT with successful reperfusion were included for final analysis. Ground truth infarct lesions were segmented on 24 h follow-up CT scans. Pre-processed CTA images were used as input for a U-Net-based convolutional neural network trained for lesion prediction, enhanced with a spatial and channel-wise squeeze-and-excitation block. Post-processing was applied to remove small predicted lesion components. The model was evaluated using a 5-fold cross-validation and a separate test set with Dice similarity coefficient (DSC) as the primary metric and average volume error as the secondary metric. The mean ± standard deviation test set DSC over all folds after post-processing was 0.35 ± 0.2 and the mean test set average volume error was 11.5 mL. The performance was relatively uniform across models with the best model according to the DSC achieved a score of 0.37 ± 0.2 after post-processing and the best model in terms of average volume error yielded 3.9 mL. 24 h follow-up infarct prediction using acute CTA imaging exclusively is feasible with DSC measures comparable to results of CTP-based algorithms reported in other studies. The proposed method might pave the way to a wider acceptance, feasibility, and applicability of follow-up infarct prediction based on artificial intelligence.

Glenoid segmentation from computed tomography scans based on a 2-stage deep learning model for glenoid bone loss evaluation.

The best-fitting circle drawn by computed tomography (CT) reconstruction of the en face view of the glenoid bone to measure the bone defect is widely used in clinical application. However, there are still some limitations in practical application, which can prevent the achievement of accurate measurements. This study aimed to accurately and automatically segment the glenoid from CT scans based on a 2-stage deep learning model and to quantitatively measure the glenoid bone defect. Patients who were referred to our institution between June 2018 and February 2022 were retrospectively reviewed. The dislocation group consisted of 237 patients with a history of ≥2 unilateral shoulder dislocations within 2 years. The control group consisted of 248 individuals with no history of shoulder dislocation, shoulder developmental deformity, or other disease that may lead to abnormal morphology of the glenoid. All patients underwent CT examination with a 1-mm slice thickness and a 1-mm increment, including complete imaging of the bilateral glenoid. A residual neural network (ResNet) location model and a U-Net bone segmentation model were constructed to develop an automated segmentation model for the glenoid from CT scans. The data set was randomly divided into training (201 of 248) and test (47 of 248) data sets of control-group data and training (190 of 237) and test (47 of 237) data sets of dislocation-group data. The accuracy of the stage 1 (glenoid location) model, the mean intersection-over-union value of the stage 2 (glenoid segmentation) model, and the glenoid volume error were used to assess the performance of the model. The R2 value and Lin concordance correlation coefficient were used to assess the correlation between the prediction and the gold standard. A total of 73,805 images were obtained after the labeling process, and each image was composed of CT images of the glenoid and its corresponding mask. The average overall accuracy of stage 1 was 99.28%; the average mean intersection-over-union value of stage 2 was 0.96. The average glenoid volume error between the predicted and true values was 9.33%. The R2 values of the predicted and true values of glenoid volume and glenoid bone loss (GBL) were 0.87 and 0.91, respectively. The Lin concordance correlation coefficient value of the predicted and true values of glenoid volume and GBL were 0.93 and 0.95, respectively. The 2-stage model in this study showed a good performance in glenoid bone segmentation from CT scans and could quantitatively measure GBL, providing a data reference for subsequent clinical treatment.

Hybrid Bladder Phantom to Validate Next-Generation Optical Wearables for Neurogenic Bladder Volume Monitoring.

The development of optics-based wearables for bladder volume monitoring has emerged as a significant topic in recent years. Given the innovative nature of this technology, there is currently no bladder phantom available to effectively validate these devices against more established gold standards, such as ultrasound. In this study, we showcase and demonstrate the performance of our hybrid bladder phantom by using an optical device and making comparisons with ultrasound. A series of validation tests, including phantom repeatability, ultrasound scanning, and an optical test, were performed. A near-infrared optical device was utilized to conduct diffuse optical spectroscopy (DOS). Machine learning models were employed to construct predictive models of volume using optical signals. The size and position of an embedded balloon, serving as an analog for the bladder, were shown to be consistent when infused with 100 mL to 350 mL of water during repeatability testing. For DOS data, we present 7 types of machine learningbased models based on different optical signals. The 2 best-performing models demonstrated an average absolute volume error ranging from 12.7 mL to 19.0 mL. In this study, we introduced a hybrid bladder phantom designed for the validation of near-infrared spectroscopy-based bladder monitoring devices in comparison with ultrasound techniques. By offering a reproducible and robust validation tool, we aim to support the advancement of next-generation optical wearables for bladder volume monitoring.

CAvity DEtection Tool (CADET): pipeline for detection of X-ray cavities in hot galactic and cluster atmospheres

ABSTRACT The study of jet-inflated X-ray cavities provides a powerful insight into the energetics of hot galactic atmospheres and radio-mechanical AGN feedback. By estimating the volumes of X-ray cavities, the total energy and thus also the corresponding mechanical jet power required for their inflation can be derived. Properly estimating their total extent is, however, non-trivial, prone to biases, nearly impossible for poor-quality data, and so far has been done manually by scientists. We present a novel machine-learning pipeline called Cavity Detection Tool (CADET), developed as an assistive tool that detects and estimates the sizes of X-ray cavities from raw Chandra images. The pipeline consists of a convolutional neural network trained for producing pixel-wise cavity predictions and a DBSCAN clustering algorithm, which decomposes the predictions into individual cavities. The convolutional network was trained using mock observations of early-type galaxies simulated to resemble real noisy Chandra-like images. The network’s performance has been tested on simulated data obtaining an average cavity volume error of 14 per cent at an 89 per cent true-positive rate. For simulated images without any X-ray cavities inserted, we obtain a 5 per cent false-positive rate. When applied to real Chandra images, the pipeline recovered 93 out of 97 previously known X-ray cavities in nearby early-type galaxies and all 14 cavities in chosen galaxy clusters. Besides that, the CADET pipeline discovered seven new cavity pairs in atmospheres of early-type galaxies (IC 4765, NGC 533, NGC 2300, NGC 3091, NGC 4073, NGC 4125, and NGC 5129) and a number of potential cavity candidates.

An End-to-End Physics-Informed Neural Network for Defect Identification and 3-D Reconstruction Using Rotating Alternating Current Field Measurement

Alternating current field measurement (ACFM) technique has been widely used in defect detection of metal structures. However, the identification and reconstruction of defects depend on human experience or simple empirical formulas, which leads to misjudgment of defects and large quantization errors. This paper proposes an end-to-end physics-informed neural network for defect identification and three-dimensional (3-D) reconstruction. The high precision automatic detection system with a specially designed probe is established to detect defects in any direction. The Faster-RCNN network is used to identify and classify defects. The physics-informed Pix2Pix network is constructed to realize 3-D reconstruction of defects with different types. The results show that the established end-to-end physics-informed neural network can realize the identification and 3-D reconstruction of defects, in which the mean average precision is 0.9982, the average length error of cracks is 0.9249 mm, the average depth error of cracks is 0.3402 mm, the average volume error of corrosion is 0.0667, and the average maximum depth error of corrosion is 0.3464 mm.

A comprehensive array of gray matter labels for the MIITRA atlas: Interoperability with complementary atlases

AbstractBackgroundA comprehensive array of gyral‐based,cytoarchitecture‐based,and functional connectivity‐based gray matter labels were constructed in the Multichannel Illinois Institute of Technology & Rush university Aging (MIITRA) atlas space in order to enhance the functionality of the atlas and its interoperability with complementary atlases.MethodT1w images from the 400 older adults included in the construction of the MIITRA atlas (50%male;64.9‐98.9 years of age) were processed with Freesurfer’s standard recon‐all pipeline. The Freesurfer outputs for all images were manually edited. Parcellations from Yeo,Desikan‐Killiany,HCP‐MMP,Brainnetome,Brodmann,Destrieux,Campbell,Flechsig,Mindboggle,Kleist,Smith,and EconomoCT atlases were projected onto the surface of each image and then converted to volumetric labels. The T1w images were also nonlinearly registered using ANTs‐SyN to the MNI152‐6thgen,MNI‐Colin,and the MNI‐2009b templates, and volumetric labels from Harvard‐Oxford,Julich,AAL3,Buckner,CoBrALab,CAREN,Striatum‐subdivision and Hippocampal‐subfields atlases were warped to the space of each image. The ANTs‐derived transformations applied on individual T1w images to build the MIITRA T1w template were used to map the corresponding gray matter labels from raw space to exact physical locations in the final MIITRA space. The performance of the MIITRA gray matter labels in segmentation of the gray matter of a separate group of 100 older adult individuals was evaluated in terms of label overlap, geometry, and dissimilarity when comparing the MIITRA labels warped to each individual’s space, to the respective reference labels from the source atlases.ResultExamples of the gray matter labels in MIITRA space are shown in Figure1. The measures of overlap between the warped and reference labels for each set of labels were generally high, with an average Dice coefficient of 0.78±0.3, average Jaccard coefficient of 0.59±0.27(Fig.2), sensitivity of 0.74±0.1, and specificity of 0.86±0.35(Fig.3). The label geometry showed a high correlation both in terms of the average volume of each label(correlation coefficient = 0.997,p‐value<10−10) and in terms of the average surface area of each label(correlation coefficient = 0.991,p‐value<10−10)(Fig.4). The values of dissimilarity were low, with an average volume error of 0.38±0.2(Fig.5).ConclusionThe gray matter labels in the MIITRA, in combination with the high‐resolution T1w template of the atlas, allowed segmentation of the gray matter of older adults that is in good agreement with the reference labels from the source atlases.

Deformable image registration based on single or multi-atlas methods for automatic muscle segmentation and the generation of augmented imaging datasets.

Muscle segmentation is a process relied upon to gather medical image-based muscle characterisation, useful in directly assessing muscle volume and geometry, that can be used as inputs to musculoskeletal modelling pipelines. Manual or semi-automatic techniques are typically employed to segment the muscles and quantify their properties, but they require significant manual labour and incur operator repeatability issues. In this study an automatic process is presented, aiming to segment all lower limb muscles from Magnetic Resonance (MR) imaging data simultaneously using three-dimensional (3D) deformable image registration (single inputs or multi-atlas). Twenty-three of the major lower limb skeletal muscles were segmented from five subjects, with an average Dice similarity coefficient of 0.72, and average absolute relative volume error (RVE) of 12.7% (average relative volume error of -2.2%) considering the optimal subject combinations. The multi-atlas approach showed slightly better accuracy (average DSC: 0.73; average RVE: 1.67%). Segmented MR imaging datasets of the lower limb are not widely available in the literature, limiting the potential of new, probabilistic methods such as deep learning to be used in the context of muscle segmentation. In this work, Non-linear deformable image registration is used to generate 69 manually checked, segmented, 3D, artificial datasets, allowing access for future studies to use these new methods, with a large amount of reliable reference data.

Intelligent Whistling System of Rail Train Based on YOLOv4 and U-Net

The whistle of the rail train is usually directly controlled by the driver. However, in long-distance transportation, there is a risk of traffic accidents due to driver fatigue or distraction. In addition, the noise pollution of the train whistle has also been criticized. In order to solve the above two problems, an intelligent whistling system for railway trains based on deep learning is proposed. The system judges whether to whistle and intelligently adjusts the volume of the whistle according to the road conditions of the train. The system consists of a road condition sensing module and a whistling decision module. The former includes the target detection model based on YOLOv4 and the semantic segmentation model based on U-Net, which can extract the key information of the road conditions ahead; the latter is to carry out logical analysis of the data after the intelligent recognition and processing and make the whistling decision. Based on the train-running data set, the intelligent whistle system model is tested. The results of this research show that the whistling accuracy of the model on the test set is 99.22%, the average volume error is 1.91 dB/time, and the Frames Per Second (FPS) is 18.7 f/s. Therefore, the intelligent whistle system model proposed in this paper has high reliability and is suitable for further development and application in actual scenes.

Development and evaluation of high‐resolution gray matter labels for the MIITRA atlas

AbstractBackgroundHigh resolution gray matter labels were constructed for the Multichannel Illinois Institute of Technology & Rush university Aging (MIITRA) atlas based on manually edited gray matter labels of the older adults included in the atlas. The performance of the new labels in labeling the gray matter of a separate group of older adults was evaluated.MethodT1w images (400 images, 1mm isotropic) from the older adults included in the MIITRA atlas (50% male; 64.9‐98.9 age‐range; 54% white, 43% black) were processed with Freesurfer’s standard recon‐all pipeline, which segmented subcortical and cortical gray matter into 84 regions according to the Desikan Killiany atlas. The Freesurfer output for all images was manually edited. The ANTs‐derived transformations applied on individual T1w images to build the MIITRA T1w template were used to map the corresponding gray matter labels from raw space to exact physical locations in the final MIITRA space. The label in a 0.5mm isotropic voxel in MIITRA space was calculated using majority voting among all the labels that were mapped to that voxel. The performance of the MIITRA gray matter labels in segmentation of the gray matter of a separate group of 100 older adult individuals was evaluated in terms of label overlap, geometry, and dissimilarity when comparing the MIITRA labels warped to each individual’s space, to the respective manually edited reference labels.ResultExamples of the MIITRA gray matter labels and corresponding confidence maps are shown in Figure 1. The measures of overlap were generally high, with an average Dice coefficient of 0.82±0.10, Jaccard coefficient of 0.70±0.13 (Fig. 2), sensitivity of 0.77±0.13, and specificity of 0.89±0.09 (Fig. 3). There was a high correlation in label geometry both in terms of the average volume (correlation coefficient=0.997, p‐value<10‐10) and in terms of the average surface area (correlation coefficient=0.988, p‐value<10‐10) (Fig. 4). The values of dissimilarity were low, with an average volume error of 0.18±0.18 and an average Hausdorff distance of 8.69±8.78 (Fig. 5).ConclusionThe MIITRA gray matter labels, in combination with the high‐resolution T1w template of the atlas, allow segmentation of the gray matter of older adults that is in good agreement with the manually‐edited Freesurfer‐based segmentation.

Testing of the Storm Water Management Model Low Impact Development Modules.

Stormwater infrastructure designers and operators rely heavily on the United States Environmental Protection Agency's Storm Water Management Model (SWMM) to simulate stormwater and wastewater infrastructure performance. Since its inception in the late 1970s, improvements and extensions have been tested and evaluated rigorously to verify the accuracy of the model. As a continuation of this progress, the main objective of this study was to quantify how accurately SWMM simulates the hydrologic activity of low impact development (LID) storm control measures. Model performance was evaluated by quantitatively comparing empirical data to model results using a multievent, multiobjective calibration method. The calibration methodology utilized the PEST software, a Parameter ESTimation tool, to determine unmeasured hydrologic parameters for SWMM's LID modules. The calibrated LID modules' Nash-Sutcliffe efficiencies averaged 0.81; average percent bias (PBIAS) -9%; average ratio of root mean square error to standard deviation of measured values 0.485; average index of agreement 0.94; and the average volume error, simulated vs. observed, was +9%. SWMM accurately predicted the timing of peak flows, but usually underestimated their magnitudes by 10%. The average volume reduction, measured outflow volume divided by inflow volume, was 48%. We had more difficulty in calibrating one study, an infiltration trench, which identified a significant limitation of the current version of the SWMM LID module; it cannot simulate lateral exfiltration of water out of the storage layers of a LID storm control measure. This limitation is especially severe for a deep LIDs, such as infiltration trenches. Nevertheless, SWMM satisfactorily simulated the hydrologic performance of eight of the nine LID practices.

Machine deep learning accurately detects endoleak after endovascular abdominal aortic aneurysm repair

ObjectiveThe objective of this study was to develop a machine deep learning algorithm for endoleak detection and measurement of aneurysm diameter, area, and volume from computed tomography angiography (CTA). MethodsDigital Imaging and Communications in Medicine files representing three-phase postoperative CTA images (N = 334) of 191 unique patients undergoing endovascular aneurysm repair for infrarenal abdominal aortic aneurysm (AAA) with a variety of commercial devices were used to train a deep learning pipeline across four tasks. The RetinaNet object-detection convolutional neural network (CNN) architecture was trained to predict bounding boxes around the axial CTA slices that were then stitched together in two dimensions into a smaller region containing the aneurysm. Multiclass endoleak detection and segmentation of the AAA, endograft, and endoleak were performed on this smaller region. Segmentations on a single randomly selected contrast from each scan included 33 full and 68 partial segmentations for endograft and AAA and 99 full segmentations for endoleak. A modified version of ResNet-50 CNN was used to detect endoleak on individual axial slices. A three-dimensional U-Net CNN model was trained on the task of dense three-dimensional segmentation and used to measure diameter and volume with a specially designed loss function. We made use of fivefold cross-validation to evaluate model performance for each step, splitting training and testing data at each fold, such that multiple scans from the same patient were preserved with the same fold. Algorithm predictions for endoleak were compared with the radiology report and with a subset of CTA images independently read by two vascular specialists. ResultsThe localization portion of the network accurately predicted a region of interest containing the AAA in 99% of cases. The best model of binary endoleak detection obtained an area under the receiver operating characteristic curve of 0.94 ± 0.03 with an optimized accuracy of 0.89 ± 0.03 on a balanced data set. An introduced postprocessing algorithm for determining maximum diameter was used on both the predicted AAA segmentation and ground truth segmentation, predicting on average an absolute diameter error of 2.3 ± 2.0 mm by 1.4 ± 1.7 mm for each measurement, respectively. The algorithm measured AAA and endograft volume accurately (Dice coefficient, 0.95 ± 0.2) with an absolute volume error of 10.1 ± 9.1 mL. The algorithm measured endoleak volume less accurately, with a Dice score of 0.53 ± 0.21 and an average absolute volume error of 1.2 ± 1.9 mL. ConclusionsThis machine learning algorithm shows promise in augmenting a human's ability to interpret postoperative CTA images and may help improve surveillance after endovascular aneurysm repair. External validation on larger data sets and prospective study are required before the algorithm can be clinically applicable.

U-Net based deep learning bladder segmentation in CT urography.

To develop a U-Net-based deep learning approach (U-DL) for bladder segmentation in computed tomography urography (CTU) as a part of a computer-assisted bladder cancer detection and treatment response assessment pipeline. A dataset of 173 cases including 81 cases in the training/validation set (42 masses, 21 with wall thickening, 18 normal bladders), and 92 cases in the test set (43 masses, 36 with wall thickening, 13 normal bladders) were used with Institutional Review Board approval. An experienced radiologist provided three-dimensional (3D) hand outlines for all cases as the reference standard. We previously developed a bladder segmentation method that used a deep learning convolution neural network and level sets (DCNN-LS) within a user-input bounding box. However, some cases with poor image quality or with advanced bladder cancer spreading into the neighboring organs caused inaccurate segmentation. We have newly developed an automated U-DL method to estimate a likelihood map of the bladder in CTU. The U-DL did not require a user-input box and the level sets for postprocessing. To identify the best model for this task, we compared the following models: (a) two-dimensional (2D) U-DL and 3D U-DL using 2D CT slices and 3D CT volumes, respectively, as input, (b) U-DLs using CT images of different resolutions as input, and (c) U-DLs with and without automated cropping of the bladder as an image preprocessing step. The segmentation accuracy relative to the reference standard was quantified by six measures: average volume intersection ratio (AVI), average percent volume error (AVE), average absolute volume error (AAVE), average minimum distance (AMD), average Hausdorff distance (AHD), and the average Jaccard index (AJI). As a baseline, the results from our previous DCNN-LS method were used. In the test set, the best 2D U-DL model achieved AVI, AVE, AAVE, AMD, AHD, and AJI values of 93.4±9.5%, -4.2±14.2%, 9.2±11.5%, 2.7±2.5mm, 9.7±7.6mm, 85.0±11.3%, respectively, while the corresponding measures by the best 3D U-DL were 90.6±11.9%, -2.3±21.7%, 11.5±18.5%, 3.1±3.2mm, 11.4±10.0mm, and 82.6±14.2%, respectively. For comparison, the corresponding values obtained with the baseline method were 81.9±12.1%, 10.2±16.2%, 14.0±13.0%, 3.6±2.0mm, 12.8±6.1mm, and 76.2±11.8%, respectively, for the same test set. The improvement for all measures between the best U-DL and the DCNN-LS were statistically significant (P<0.001). Compared to a previous DCNN-LS method, which depended on a user-input bounding box, the U-DL provided more accurate bladder segmentation and was more automated than the previous approach.

Deep-learning convolutional neural network: Inner and outer bladder wall segmentation in CT urography.

We are developing a computerized segmentation tool for the inner and outer bladder wall as a part of an image analysis pipeline for CT urography (CTU). A data set of 172 CTU cases was collected retrospectively with Institutional Review Board (IRB) approval. The data set was randomly split into two independent sets of training (81 cases) and testing (92 cases) which were manually outlined for both the inner and outer wall. We trained a deep-learning convolutional neural network (DL-CNN) to distinguish the bladder wall from the inside and outside of the bladder using neighborhood information. Approximately, 240000 regions of interest (ROIs) of 16×16 pixels in size were extracted from regions in the training cases identified by the manually outlined inner and outer bladder walls to form a training set for the DL-CNN; half of the ROIs were selected to include the bladder wall and the other half were selected to exclude the bladder wall with some of these ROIs being inside the bladder and the rest outside the bladder entirely. The DL-CNN trained on these ROIs was applied to the cases in the test set slice-by-slice to generate a bladder wall likelihood map where the gray level of a given pixel represents the likelihood that a given pixel would belong to the bladder wall. We then used the DL-CNN likelihood map as an energy term in the energy equationof a cascaded level sets method to segment the inner and outer bladder wall. The DL-CNN segmentation with level sets was compared to the three-dimensional (3D) hand-segmented contours as a reference standard. For the inner wall contour, the training set achieved the average volume intersection, average volume error, average absolute volume error, and average distance of 90.0±8.7%, -4.2±18.4%, 12.9±13.9%, and 3.0±1.6mm, respectively. The corresponding values for the test set were 86.9±9.6%, -8.3±37.7%, 18.4±33.8%, and 3.4±1.8mm, respectively. For the outer wall contour, the training set achieved the values of 93.7±3.9%, -7.8±11.4%, 10.3±9.3%, and 3.0±1.2mm, respectively. The corresponding values for the test set were 87.5±9.9%, -1.2±20.8%, 11.9±17.0%, and 3.5±2.3mm, respectively. Our study demonstrates that DL-CNN-assisted level sets can effectively segment bladder walls from the inner bladder and outer structures despite a lack of consistent distinctions along the inner wall. However, even with the addition of level sets, the inner and outer walls may still be over-segmented and the DL-CNN-assisted level sets may incorrectly segment parts of the prostate that overlap with the outer bladder wall. The outer wall segmentation was improved compared to our previous method and the DL-CNN-assisted level sets were also able to segment the inner bladder wall with similar performance. This study shows the DL-CNN-assisted level set segmentation tool can effectively segment the inner and outer wall of the bladder.

Factors controlling alterations in the performance of a runoff model in changing climate conditions

Abstract In many Austrian catchments in recent decades an increase in the mean annual air temperature and precipitation has been observed, but only a small change in the mean annual runoff. The main objective of this paper is (1) to analyze alterations in the performance of a conceptual hydrological model when applied in changing climate conditions and (2) to assess the factors and model parameters that control these changes. A conceptual rainfall-runoff model (the TUW model) was calibrated and validated in 213 Austrian basins from 1981–2010. The changes in the runoff model’s efficiency have been compared with changes in the mean annual precipitation and air temperature and stratified for basins with dominant snowmelt and soil moisture processes. The results indicate that while the model’s efficiency in the calibration period has not changed over the decades, the values of the model’s parameters and hence the model’s performance (i.e., the volume error and the runoff model’s efficiency) in the validation period have changed. The changes in the model’s performance are greater in basins with a dominant soil moisture regime. For these basins, the average volume error which was not used in calibration has increased from 0% (in the calibration periods 1981–1990 or 2001–2010) to 9% (validation period 2001–2010) or –8% (validation period 1981–1990), respectively. In the snow-dominated basins, the model tends to slightly underestimate runoff volumes during its calibration (average volume error = –4%), but the changes in the validation periods are very small (i.e., the changes in the volume error are typically less than 1–2%). The model calibrated in a colder decade (e.g., 1981–1990) tends to overestimate the runoff in a warmer and wetter decade (e.g., 2001–2010), particularly in flatland basins. The opposite case (i.e., the use of parameters calibrated in a warmer decade for a colder, drier decade) indicates a tendency to underestimate runoff. A multidimensional analysis by regression trees showed that the change in the simulated runoff volume is clearly related to the change in precipitation, but the relationship is not linear in flatland basins. The main controlling factor of changes in simulated runoff volumes is the magnitude of the change in precipitation for both groups of basins. For basins with a dominant snowmelt runoff regime, the controlling factors are also the wetness of the basins and the mean annual precipitation. For basins with a soil moisture regime, landcover (forest) plays an important role.

Using Multiple Monthly Water Balance Models to Evaluate Gridded Precipitation Products over Peninsular Spain

The availability of precipitation data is the key driver in the application of hydrological models when simulating streamflow. Ground weather stations are regularly used to measure precipitation. However, spatial coverage is often limited in low-population areas and mountain areas. To overcome this limitation, gridded datasets from remote sensing have been widely used. This study evaluates four widely used global precipitation datasets (GPDs): The Tropical Rainfall Measuring Mission (TRMM) 3B43, the Climate Forecast System Reanalysis (CFSR), the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), and the Multi-Source Weighted-Ensemble Precipitation (MSWEP), against point gauge and gridded dataset observations using multiple monthly water balance models (MWBMs) in four different meso-scale basins that cover the main climatic zones of Peninsular Spain. The volumes of precipitation obtained from the GPDs tend to be smaller than those from the gauged data. Results underscore the superiority of the national gridded dataset, although the TRMM provides satisfactory results in simulating streamflow, reaching similar Nash-Sutcliffe values, between 0.70 and 0.95, and an average total volume error of 12% when using the GR2M model. The performance of GPDs highly depends on the climate, so that the more humid the watershed is, the better results can be achieved. The procedures used can be applied in regions with similar case studies to more accurately assess the resources within a system in which there is scarcity of recorded data available.

Improving The Accuracy Of A Turbine Spirometer At Low Flow Rates

The turbine spirometer is a popular laboratory device used to measure respiratory volumes and flows during exercise, and while performing gross, voluntary respiratory manoeuvres (i.e., inspiratory capacity efforts). Indeed, these devices have been incorporated into many commercially-available pulmonary function and metabolic systems. Yet, while the turbine spirometer may provide accurate/reliable measurements of respiratory volumes at modest-to-high flows, these devices perform poorly at low flow rates. PURPOSE: To improve the accuracy of a turbine spirometer over an extended range of low flows using the “weighted averaging technique” described by Yeh et al. (J Appl Physiol, 53(1): p280, 1982). METHODS: A commercially-available turbine spirometer was interfaced with a custom-designed microcontroller unit (MCU). The MCU recorded discrete rotations of the turbine rotor, and the corresponding rotational frequency (frot). Repeated 5-fold cross-validation was used to determine the optimal number of bins in frot and iterations used in the Weighted Averaging algorithm. This method yielded a discrete array of calibration constants (K) across a relevant range of frot. (<1-1800 Hz). The accuracy of this “nonlinear” calibration curve was compared to that obtained by assuming a constant K across all frequencies (i.e., flows). Over 200 calibrations strokes were recorded using a 3 L syringe. RESULTS: By assuming a constant K (15.6 mL[BULLET OPERATOR]pulse-1), the turbine spirometer exhibited an average volume error of +94 mL (+3.1%) over a 95% confidence interval (CI95%) of -856 to +375 mL (-28.5 to 12.5%). Conversely, applying the nonlinear K curve resulted in an average volume error of <0.001 mL (<0.001%) and a CI95% ranging from -60 to +60 mL (-2.0 to 2.0%). Importantly, the nonlinear K curve provided accurate (within ±3%) volume measurements down to 0.33 Hz (~7 mL[BULLET OPERATOR]s-1). CONCLUSIONS: The “weighted averaging technique” improved the accuracy/reliability the turbine spirometer to within ±3% across an interval of flows ranging between ~0.01 to 20 L[BULLET OPERATOR]s-1.

Gyri of the human parietal lobe: Volumes, spatial extents, automatic labelling, and probabilistic atlases.

Accurately describing the anatomy of individual brains enables interlaboratory communication of functional and developmental studies and is crucial for possible surgical interventions. The human parietal lobe participates in multimodal sensory integration including language processing and also contains the primary somatosensory area. We describe detailed protocols to subdivide the parietal lobe, analyze morphological and volumetric characteristics, and create probabilistic atlases in MNI152 stereotaxic space. The parietal lobe was manually delineated on 3D T1 MR images of 30 healthy subjects and divided into four regions: supramarginal gyrus (SMG), angular gyrus (AG), superior parietal lobe (supPL) and postcentral gyrus (postCG). There was the expected correlation of male gender with larger brain and intracranial volume. We examined a wide range of anatomical features of the gyri and the sulci separating them. At least a rudimentary primary intermediate sulcus of Jensen (PISJ) separating SMG and AG was identified in nearly all (59/60) hemispheres. Presence of additional gyri in SMG and AG was related to sulcal features and volumetric characteristics. The parietal lobe was slightly (2%) larger on the left, driven by leftward asymmetries of the postCG and SMG. Intersubject variability was highest for SMG and AG, and lowest for postCG. Overall the morphological characteristics tended to be symmetrical, and volumes also tended to covary between hemispheres. This may reflect developmental as well as maturation factors. To assess the accuracy with which the labels can be used to segment newly acquired (unlabelled) T1-weighted brain images, we applied multi-atlas label propagation software (MAPER) in a leave-one-out experiment and compared the resulting automatic labels with the manually prepared ones. The results showed strong agreement (mean Jaccard index 0.69, corresponding to a mean Dice index of 0.82, average mean volume error of 0.6%). Stereotaxic probabilistic atlases of each subregion were obtained. They illustrate the physiological brain torque, with structures in the right hemisphere positioned more anteriorly than in the left, and right/left positional differences of up to 10 mm. They also allow an assessment of sulcal variability, e.g. low variability for parietooccipital fissure and cingulate sulcus. Illustrated protocols, individual label sets, probabilistic atlases, and a maximum-probability atlas which takes into account surrounding structures are available for free download under academic licences.

Macroanatomy and 3D probabilistic atlas of the human insula

The human insula is implicated in numerous functions. More and more neuroimaging studies focus on this region, however no atlas offers a complete subdivision of the insula in a reference space. The aims of this study were to define a protocol to subdivide insula, to create probability maps in the MNI152 stereotaxic space, and to provide normative reference volume measurements for these subdivisions.Six regions were manually delineated bilaterally on 3D T1 MR images of 30 healthy subjects: the three short gyri, the anterior inferior cortex, and the two long gyri.The volume of the insular grey matter was 7.7 ± 0.9cm3 in native space and 9.9 ± 0.6cm3 in MNI152 space. These volumes expressed as a percentage of the ipsilateral grey matter volume were minimally larger in women (2.7±0.2%) than in men (2.6±0.2%). After spatial normalization, a stereotactic probabilistic atlas of each subregion was produced, as well as a maximum-probability atlas taking into account surrounding structures.Automatically labelling insular subregions via a multi-atlas propagation and label fusion strategy (MAPER) in a leave-one-out experiment showed high spatial overlaps of such automatically defined insular subregions with the manually derived ones (mean Jaccard index 0.65, corresponding to a mean Dice index of 0.79), with an average mean volume error of 2.6%.Probabilistic and maximum probability atlases and the original delineations are available on the web under free academic licences.

SU‐F‐T‐544: Development of a Clinically Feasible Workflow of 4D MRI‐Based Treatment Planning for Liver SBRT

Purpose:To develop a clinically feasible workflow of 4D MRI‐based treatment planning for liver SBRT.Methods:A clinical workflow based on 4D MRI for liver SBRT was designed to largely simulate the same procedure as 4D CT for lung SBRT. Key features of the workflow include: (1) contouring tumor volume at each phases (2) contouring more structures, (3) assigning CT numbers to structures for dose calculation. Uncertainties (bone delineation, MRI distortion, CT number assignment, etc.) were investigated using five clinical treatment plans of liver SBRT by simulating the following scenarios: homogeneous calculation, calculation with CT number assignment (for each structure separately and for all combined), calculation with and without distortion (mimicked by shirking body structure by 1mm to 5 mm). Dose‐histogram volume (DVF) parameters were collected from the simulated cases and compared to those in the original plan. Lastly, a patient example case of 4D MRI‐based treatment plan was performed to demonstrate the feasibility of the proposed workflow.Results:In homogeneous calculation, the average percentage volume error (ΔV30) of liver was 0.40% and maximum average dose error (GTV ΔD95) was 1.08Gy. Between MRI‐based and original plans, liver ΔV30 was 0.28% and maximum average dose error (GTV ΔD95) was 0.75Gy. Basically dose errors and percentage volume error increased as distortion increased. In some special cases, in original plans, the minimum spinal cord max dose error appeared when distortion was 4mm; in MRI‐based plans, the minimum liver ΔV30 appeared when distortion was 3mm. In the example case, mean liver dose was 20.7Gy in 4D MRI‐based plan and 34.2Gy in 4D CT‐based plan.Conclusion:A clinically feasible workflow of 4D MRI‐based treatment planning for liver SBRT has been established. Dose uncertainties of the workflow were generally small, except for large MR geometric distortions (&gt; 3 mm).NIH (1R21CA165384)