Fully Automated Detection of the Appendix Using U-Net Deep Learning Architecture in CT Scans

Despite extensive efforts to obtain accurate segmentation of magnetic resonance imaging (MRI) scans of a head, it remains challenging primarily due to variations in intensity distribution, which depend on the equipment and parameters used. The goal of this study is to evaluate the effectiveness of an automatic segmentation method for head MRI scans using a multistep Dense U-Net (MDU-Net) architecture. The MDU-Net-based method comprises two steps. The first step is to segment the scalp, skull, and whole brain from head MRI scans using a convolutional neural network (CNN). In the first step, a hybrid network is used to combine 2.5D Dense U-Net and 3D Dense U-Net structure. This hybrid network acquires logits in three orthogonal planes (axial, coronal, and sagittal) using 2.5D Dense U-Nets and fuses them by averaging. The resultant fused probability map with head MRI scans then serves as the input to a 3D Dense U-Net. In this process, different ratios of active contour loss and focal loss are applied. The second step is to segment the cerebrospinal fluid (CSF), white matter, and gray matter from extracted brain MRI scans using CNNs. In the second step, the histogram of the extracted brain MRI scans is standardized and then a 2.5D Dense U-Net is used to further segment the brain's specific tissues using the focal loss. A dataset of 100 head MRI scans from an OASIS-3 dataset was used for training, internal validation, and testing, with ratios of 80%, 10%, and 10%, respectively. Using the proposed approach, we segmented the head MRI scans into five areas (scalp, skull, CSF, white matter, and gray matter) and evaluated the segmentation results using the Dice similarity coefficient (DSC) score, Hausdorff distance (HD), and the average symmetric surface distance (ASSD) as evaluation metrics. We compared these results with those obtained using the Res-U-Net, Dense U-Net, U-Net++, Swin-Unet, and H-Dense U-Net models. The MDU-Net model showed DSC values of 0.933, 0.830, 0.833, 0.953, and 0.917 in the scalp, skull, CSF, white matter, and gray matter, respectively. The corresponding HD values were 2.37, 2.89, 2.13, 1.52, and 1.53mm, respectively. The ASSD values were 0.50, 1.63, 1.28, 0.26, and 0.27mm, respectively. Comparing these results with other models revealed that the MDU-Net model demonstrated the best performance in terms of the DSC values for the scalp, CSF, white matter, and gray matter. When compared with the H-Dense U-Net model, which showed the highest performance among the other models, the MDU-Net model showed substantial improvements in the HD view, particularly in the gray matter region, with a difference of approximately 9%. In addition, in terms of the ASSD, the MDU-Net model outperformed the H-Dense U-Net model, showing an approximately 7% improvements in the white matter and approximately 9% improvements in the gray matter. Compared with existing models in terms of DSC, HD, and ASSD, the proposed MDU-Net model demonstrated the best performance on average and showed its potential to enhance the accuracy of automatic segmentation for head MRI scans.

A multi-institute automated segmentation evaluation on a standard dataset: Findings from the international workshop on osteoarthritis imaging segmentation challenge

We propose a method for automatic segmentation of individual muscles from a clinical CT. The method uses Bayesian convolutional neural networks with the U-Net architecture, using Monte Carlo dropout that infers an uncertainty metric in addition to the segmentation label. We evaluated the performance of the proposed method using two data sets: 20 fully annotated CTs of the hip and thigh regions and 18 partially annotated CTs that are publicly available from The Cancer Imaging Archive (TCIA) database. The experiments showed a Dice coefficient (DC) of 0.891±0.016 (mean±std) and an average symmetric surface distance (ASD) of 0.994±0.230 mm over 19 muscles in the set of 20 CTs. These results were statistically significant improvements compared to the state-of-the-art hierarchical multi-atlas method which resulted in 0.845 ± 0.031 DC and 1.556 ± 0.444 mm ASD. We evaluated validity of the uncertainty metric in the multi-class organ segmentation problem and demonstrated a correlation between the pixels with high uncertainty and the segmentation failure. One application of the uncertainty metric in active-learning is demonstrated, and the proposed query pixel selection method considerably reduced the manual annotation cost for expanding the training data set. The proposed method allows an accurate patient-specific analysis of individual muscle shapes in a clinical routine. This would open up various applications including personalization of biomechanical simulation and quantitative evaluation of muscle atrophy.

This study aimed to develop a deep learning-based method for proximal humeral vBMD assessment and to investigate its age- and sex-related changes, as well as its correlation with lumbar spine vBMD. An nnU-Net-based deep learning pipeline was developed to automatically segment the proximal humerus on chest CT scans from 2,675 adults. Segmentation performance was assessed using the Dice Similarity Coefficient (DSC), Intersection over Union (IoU), 95th-percentile Hausdorff Distance (95HD), and Average Symmetric Surface Distance (ASSD). Phantom-calibrated vBMD-total, trabecular, and BMAT-corrected trabecular-was quantified for each subject. Age-related distributions were modeled with generalized additive models for location, scale, and shape (GAMLSS) to generate sex-specific P3-P97 percentile curves. Lumbar spine vBMD was measured in 1460 individuals for correlation analysis. Segmentation was highly accurate (DSC 98.42 ± 0.20%; IoU 96.89 ± 0.42%; 95HD 1.12 ± 0.37mm; ASSD 0.94 ± 0.31mm). In males, total, trabecular, and BMAT-corrected trabecular vBMD declined approximately linearly from early adulthood. In females, a pronounced inflection occurred at ~ 40-45years: values were stable or slightly rising beforehand, then all percentiles dropped steeply and synchronously, indicating accelerated menopause-related loss. In females, vBMD declined earlier in the lumbar spine than in the proximal humerus. Correlations between proximal humeral and lumbar spine vBMD were low to moderate overall and weakened after age 50. We present a novel, automated method for quantifying proximal humeral vBMD from chest CT, revealing distinct, sex-specific aging patterns. Males' humeral vBMD declines linearly, while females experience an earlier, accelerated loss. Moreover, the peak humeral vBMD in females occurs later than that of the lumbar spine, and spinal measurements cannot reliably substitute for humeral BMD in clinical assessment.

Accurate segmentation of pancreatic ductal adenocarcinoma (PDAC) and surrounding anatomical structures is critical for diagnosis, treatment planning, and outcome assessment. This study proposes a deep learning-based framework to automate multi-class segmentation in CT images, comparing the performance of four state-of-the-art architectures. This retrospective multi-center study included 3265 patients from six institutions. Four deep learning models-UNet, nnU-Net, UNETR, and Swin-UNet-were trained using five-fold cross-validation on data from five centers and tested independently on a sixth center (n = 569). Preprocessing included intensity normalization, voxel resampling, and standardized annotation for six structures: PDAC lesion, pancreas, veins, arteries, pancreatic duct, and common bile duct. Evaluation metrics included Dice Similarity Coefficient (DSC), Intersection over Union (IoU), directed Hausdorff Distance (dHD), Average Symmetric Surface Distance (ASSD), and Volume Overlap Error (VOE). Statistical comparisons were made using Wilcoxon signed-rank tests with Bonferroni correction. Swin-UNet outperformed all models with a mean validation DSC of 92.4% and test DSC of 90.8%, showing minimal overfitting. It also achieved the lowest dHD (4.3mm), ASSD (1.2mm), and VOE (6.0%) in cross-validation. Per-class DSCs for Swin-UNet were consistently higher across all anatomical targets, including challenging structures like the pancreatic duct (91.0%) and bile duct (91.8%). Statistical analysis confirmed the superiority of Swin-UNet (p < 0.001). All models showed generalization capability, but Swin-UNet provided the most accurate and robust segmentation across datasets. Transformer-based architectures, particularly Swin-UNet, enable precise and generalizable multi-class segmentation of PDAC and surrounding anatomy. This framework has potential for clinical integration in PDAC diagnosis, staging, and therapy planning.

Automatic primary gross tumor volume (GTVp) segmentation for nasopharyngeal carcinoma (NPC) is a quite challenging task because of the existence of similar visual characteristics between tumors and their surroundings, especially on computed tomography (CT) images with severe low contrast resolution. Therefore, most recently proposed methods based on radiomics or deep learning (DL) is difficult to achieve good results on CT datasets. A peritumoral radiomics-guided generative adversarial network (PRG-GAN) was proposed to address this challenge. A total of 157 NPC patients with CT images was collected and divided into training, validation, and testing cohorts of 108, 9, and 30 patients, respectively. The proposed model was based on a standard GAN consisting of a generator network and a discriminator network. Morphological dilation on the initial segmentation results from GAN was first conducted to delineate annular peritumoral region, in which radiomics features were extracted as priori guide knowledge. Then, radiomics features were fused with semantic features by the discriminator's fully connected layer to achieve the voxel-level classification and segmentation. The dice similarity coefficient (DSC), 95% Hausdorff distance (HD95), and average symmetric surface distance (ASSD) were used to evaluate the segmentation performance using a paired samples t-test with Bonferroni correction and Cohen's d (d) effect sizes. A two-sided p-value of less than 0.05 was considered statistically significant. The model-generated predictions had a high overlap ratio with the ground truth. The average DSC, HD95, and ASSD were significantly improved from 0.80±0.12, 4.65±4.71mm, and 1.35±1.15mm of GAN to 0.85±0.18 (p=0.001, d=0.71), 4.15±7.56mm (p=0.002, d=0.67), and 1.11±1.65mm (p<0.001, d=0.46) of PRG-GAN, respectively. Integrating radiomics features into GAN is promising to solve unclear border limitations and increase the delineation accuracy of GTVp for patients with NPC.

PurposeRespiratory motion of locally advanced non-small cell lung cancer (LA-NSCLC) adds to the challenge of targeting the disease with radiotherapy (RT). One technique used frequently to alleviate this challenge is an internal gross tumor volume (IGTV) generated from manual contours on a single respiratory phase of the 4DCT via the aid of deformable image registration (DIR)-based auto-propagation. Through assessing the accuracy of DIR-based auto-propagation for generating IGTVs, this study aimed to identify erring characteristics associated with the process to enhance RT targeting in LA-NSCLC.Methods4DCTs of 19 patients with LA-NSCLC were acquired using retrospective gating with 10 respiratory phases (RPs). Ground-truth IGTVs (GT-IGTVs) were obtained through manual segmentation and union of gross tumor volumes (GTVs) in all 10 phases. IGTV auto-propagation was carried out using two distinct DIR algorithms for the manually contoured GTV from each of the 10 phases, resulting in 10 separate IGTVs for each patient per each algorithm. Differences between the auto-propagated IGTVs (AP-IGTVs) and their corresponding GT-IGTVs were assessed using Dice coefficient (DICE), maximum symmetric surface distance (MSSD), average symmetric surface distance (ASSD), and percent volume difference (PVD) and further examined in relation to anatomical tumor location, RP, and deformation index (DI) that measures the degree of deformation during auto-propagation. Furthermore, dosimetric implications due to the analyzed differences between the AP-IGTVs and GT-IGTVs were assessed.ResultsFindings were largely consistent between the two algorithms: DICE, MSSD, ASSD, and PVD showed no significant differences between the 10 RPs used for propagation (Kruskal–Wallis test, ps > 0.90); MSSD and ASSD differed significantly by tumor location in the central–peripheral and superior–inferior dimensions (ps < 0.0001) while only in the central–peripheral dimension for PVD (p < 0.001); DICE, MSSD, and ASSD significantly correlated with the DI (Spearman’s rank correlation test, ps < 0.0001). Dosimetric assessment demonstrated that 79% of the radiotherapy plans created by targeting planning target volumes (PTVs) derived from the AP-IGTVs failed prescription constraints for their corresponding ground-truth PTVs.ConclusionIn LA-NSCLC, errors in DIR-based IGTV propagation present to varying degrees and manifest dependences on DI and anatomical tumor location, indicating the need for personalized consideration in designing RT internal target volume.

Radiological follow-up of oncology patients requires the detection of metastatic lung lesions and the quantitative analysis of their changes in longitudinal imaging studies. Our aim was to evaluate SimU-Net, a novel deep learning method for the automatic analysis of metastatic lung lesions and their temporal changes in pairs of chest CT scans. SimU-Net is a simultaneous multichannel 3D U-Net model trained on pairs of registered prior and current scans of a patient. It is part of a fully automatic pipeline for the detection, segmentation, matching, and classification of metastatic lung lesions in longitudinal chest CT scans. A data set of 5040 metastatic lung lesions in 344 pairs of 208 prior and current chest CT scans from 79 patients was used for training/validation (173 scans, 65 patients) and testing (35 scans, 14 patients) of a standalone 3D U-Net models and 3 simultaneous SimU-Net models. Outcome measures were the lesion detection and segmentation precision, recall, Dice score, average symmetric surface distance (ASSD), lesion matching, and classification of lesion changes from computed versus manual ground-truth annotations by an expert radiologist. SimU-Net achieved a mean lesion detection recall and precision of 0.93±0.13 and 0.79±0.24 and a mean lesion segmentation Dice and ASSD of 0.84±0.09 and 0.33±0.22mm. These results outperformed the standalone 3D U-Net model by 9.4% in the recall, 2.4% in Dice, and 15.4% in ASSD, with a minor 3.6% decrease in precision. The SimU-Net pipeline achieved perfect precision and recall (1.0±0.0) for lesion matching and classification of lesion changes. Simultaneous deep learning analysis of metastatic lung lesions in prior and current chest CT scans with SimU-Net yields superior accuracy compared with individual analysis of each scan. Implementation of SimU-Net in the radiological workflow may enhance efficiency by automatically computing key metrics used to evaluate metastatic lung lesions and their temporal changes.

This study proposes a framework to address the problem of multimodal MRI-to-CT image registration by incorporating feature-based registration approach and segmentation, focusing especially on liver-specific clinical applications. The proposed framework consists of three key stages: (1) liver segmentation using pretrained nnU-Net models for both CT and MRI images, (2) CoLlAGe features extraction within the segmented liver regions to preserve spatial information of the region of interest (ROI), and (3) image registration based on these extracted features. Across 24 pairs of registration from MRI to CT, most of the Average Symmetric Surface Distance (ASSD) results are close to 0 millimeters, and all Dice coefficients are greater than 0.8. The values of mean ± standard deviation show that the Dice coefficient is 0.921 ± 0.038, the ASSD is 0.086 ± 0.213mm. These results demonstrate the potential of the proposed framework in the study of multimodal MR-to-CT registration. The precision of liver MR-to-CT registration is considered acceptable for liver surgical applications.

Segmentation of cardiac structures is essential for cardiac function evaluation using cine magnetic resonance imaging (MRI). Deep learning can be used to segment cardiac structures in cine cardiac MRI with high accuracy, but this approach requires fully annotated datasets for training, which are difficult to obtain. Semi-supervised segmentation methods provide a way to alleviate the burden of manual labeling by using labeled and unlabeled data for training. However, these methods generally provide suboptimal segmentationaccuracies. To develop a semi-supervised method that utilizes relatively small training datasets and under-annotations for improved cine cardiac MRIsegmentation. The proposed approach consists of deformable registration, fully and weakly supervised segmentation, and a temporal attention perceiver (TAP). The registration module was trained to warp labeled frames to generate pseudo labels for unlabeled frames. The warped labeled images were used to train the fully supervised segmentation network. The unlabeled images and the pseudo label were used to train the weakly supervised segmentation model, and the segmentation prediction was compared with the input pseudo label as an auxiliary loss to the registration module. The TAP module was employed to generate optimized features for the warped labeled and the original unlabeled images both paired with the original labeled image. Consistency between the resulting features was enforced to refine cross-instance feature alignment to facilitate the registration. One hundred, twenty, and ten subjects from the Automatic Cardiac Diagnosis Challenge (ACDC) and seventy-five, thirty, and fifteen cases from the Multi-Vendor & Multi-Disease (M&Ms) Cardiac Image Segmentation Challenge were used for training, each with random end-systolic (ES)/end-diastolic (ED) frames labeled. The optimized models were used to segment the remaining 50 ACDC and 50 M&Ms subjects. The proposed approach was compared with several commonly used semi-supervised segmentation methods in terms of Dice-similarity-coefficients (DSC), average-symmetric-surface-distance (ASSD), and Hausdorff-distance (HD) for left (LV) and right (RV) ventricular cavity and myocardium (Myo). A Unet trained on the same subjects each with both frames labeled was used as an upper bound (Unet_UB). Using 100 ACDC training subjects, our approach yielded DSC = 0.910 0.063, ASSD = 1.37 0.63 mm, and HD = 6.38 2.99 mm for RV, DSC = 0.894 0.024, ASSD = 1.20 1.12 mm, and HD = 4.67 3.22 mm for Myo, and DSC = 0.934 0.056, ASSD = 1.25 1.63 mm, and HD = 3.97 5.76 mm for LV. A bidirectional copy-paste (BCP) method performed the best among the comparative methods and generated DSC = 0.902 0.060, ASSD = 1.45 0.60 mm, and HD = 7.50 3.20 mm for RV, DSC = 0.885 0.030, ASSD = 1.28 0.80 mm, and HD = 5.80 2.80 mm for Myo, and DSC = 0.920 0.068, ASSD = 1.15 0.40 mm, and HD = 4.20 3.30 mm for LV. For Unet_UB, these were 0.905 0.068, 1.48 0.61 mm, and 6.35 2.85 mm for RV, 0.895 0.030, 1.05 0.45 mm, and 4.40 3.09 mm for Myo, and 0.941 0.044, 1.02 0.34 mm, and 3.17 1.63 mm for LV. Similar trends were observed when using 75 M&Ms training subjects. For all the experiments, our approach outperformed BCP in general and yielded segmentation accuracies comparable to Unet_UB. The proposed approach outperformed several commonly used semi-supervised segmentation methods and yielded segmentation accuracies on par with fully supervised Unet using various relatively small datasets and under annotations fortraining.

Task automation is essential for efficient and consistent image segmentation in radiation oncology. We report on a deep learning architecture, comprising a U-Net and a variational autoencoder (VAE) for automatic contouring of the prostate gland incorporating interobserver variation for radiotherapy treatment planning. The U-Net/VAE generates an ensemble set of segmentations for each image CT slice. A novel outlier mitigation (OM) technique was implemented to enhance the model segmentation accuracy. The primary source dataset (source_prim) consisted of 19 200 CT slices (from 300 patient planning CT image datasets) with manually contoured prostate glands. A smaller secondary source dataset (source_sec) comprised 640 CT slices (from 10 patient CT datasets), where prostate glands were segmented by 5 independent physicians on each dataset to account for interobserver variability. Data augmentation via random rotation (<5 degrees), cropping, and horizontal flipping was applied to each dataset to increase sample size by a factor of 100. A probabilistic hierarchical U-Net with VAE was implemented and pretrained using the augmented source_prim dataset for 30 epochs. Model parameters of the U-Net/VAE were fine-tuned using the augmented source_sec dataset for 100 epochs. After the first round of training, outlier contours in the training dataset were automatically detected and replaced by the most accurate contours (based on Dice similarity coefficient, DSC) generated by the model. The U-Net/OM-VAE was retrained using the revised training dataset. Metrics for comparison included DSC, Hausdorff distance (HD, mm), normalized cross-correlation (NCC) coefficient, and center-of-mass (COM) distance (mm). Results for U-Net/OM-VAE with outliers replaced in the training dataset versus U-Net/VAE without OM were as follows: DSC=0.82±0.01 versus 0.80±0.02 (p=0.019), HD=9.18±1.22 versus 10.18±1.35mm (p=0.043), NCC=0.59±0.07 versus 0.62±0.06, and COM=3.36±0.81 versus 4.77±0.96mm over the average of 15 contours. For the average of 15 highest accuracy contours, values were as follows: DSC=0.90±0.02 versus 0.85±0.02, HD=5.47±0.02 versus 7.54±1.36mm, and COM=1.03±0.58 versus 1.46±0.68mm (p<0.03 for all metrics). Results for the U-Net/OM-VAE with outliers removed were as follows: DSC=0.78±0.01, HD=10.65±1.95mm, NCC=0.46±0.10, COM=4.17±0.79mm for the average of 15 contours, and DSC=0.88±0.02, HD=7.00±1.17mm, COM=1.58±0.63mm for the average of 15 highest accuracy contours. All metrics for U-Net/VAE trained on the source_prim and source_sec datasets via pretraining, followed by fine-tuning, show statistically significant improvement over that trained on the source_sec dataset only. Finally, all metrics for U-Net/VAE with or without OM showed statistically significant improvement over those for the standard U-Net. A VAE combined with a hierarchical U-Net and an OM strategy (U-Net/OM-VAE) demonstrates promise toward capturing interobserver variability and produces accurate prostate auto-contours for radiotherapy planning. The availability of multiple contours for each CT slice enables clinicians to determine trade-offs in selecting the "best fitting" contour on each CT slice. Mitigation of outlier contours in the training dataset improves prediction accuracy, but one must be wary of reduction in variability in the training dataset.

Automated Segmentation of Sacral Chordoma and Surrounding Muscles Using Deep Learning Ensemble

Segmentation of livers and liver tumors is one of the most important steps in radiation therapy of hepatocellular carcinoma. The segmentation task is often done manually, making it tedious, labor intensive, and subject to intra-/inter- operator variations. While various algorithms for delineating organ-at-risks (OARs) and tumor targets have been proposed, automatic segmentation of livers and liver tumors remains intractable due to their low tissue contrast with respect to the surrounding organs and their deformable shape in CT images. The U-Net has gained increasing popularity recently for image analysis tasks and has shown promising results. Conventional U-Net architectures, however, suffer from three major drawbacks. First, skip connections allow for the duplicated transfer of low resolution information in feature maps to improve efficiency in learning, but this often leads to blurring of extracted image features. Secondly, high level features extracted by the network often do not contain enough high resolution edge information of the input, leading to greater uncertainty where high resolution edge dominantly affects the network's decisions such as liver and liver-tumor segmentation. Thirdly, it is generally difficult to optimize the number of pooling operations in order to extract high level global features, since the number of pooling operations used depends on the object size. To cope with these problems, we added a residual path with deconvolution and activation operations to the skip connection of the U-Net to avoid duplication of low resolution information of features. In the case of small object inputs, features in the skip connection are not incorporated with features in the residual path. Furthermore, the proposed architecture has additional convolution layers in the skip connection in order to extract high level global features of small object inputs as well as high level features of high resolution edge information of large object inputs. Efficacy of the modified U-Net (mU-Net) was demonstrated using the public dataset of Liver tumor segmentation (LiTS) challenge 2017. For liver-tumor segmentation, Dice similarity coefficient (DSC) of 89.72 %, volume of error (VOE) of 21.93 %, and relative volume difference (RVD) of - 0.49 % were obtained. For liver segmentation, DSC of 98.51 %, VOE of 3.07 %, and RVD of 0.26 % were calculated. For the public 3D Image Reconstruction for Comparison of Algorithm Database (3Dircadb), DSCs were 96.01 % for the liver and 68.14 % for liver-tumor segmentations, respectively. The proposed mU-Net outperformed existing state-of-art networks.

To develop and validate a modified deep learning (DL) model based on nnU-Net for classifying and segmenting five-class jaw lesions using cone-beam CT (CBCT). A total of 368 CBCT scans (37 168 slices) were used to train a multi-class segmentation model. The data underwent manual annotation by two oral and maxillofacial surgeons (OMSs) to serve as ground truth. Sensitivity, specificity, precision, F1-score, and accuracy were used to evaluate the classification ability of the model and doctors, with or without artificial intelligence assistance. The dice similarity coefficient (DSC), average symmetric surface distance (ASSD), and segmentation time were used to evaluate the segmentation effect of the model. The model achieved the dual task of classifying and segmenting jaw lesions in CBCT. For classification, the sensitivity, specificity, precision, and accuracy of the model were 0.871, 0.974, 0.874, and 0.891, respectively, surpassing oral and maxillofacial radiologists (OMFRs) and OMSs, approaching the specialist. With the model's assistance, the classification performance of OMFRs and OMSs improved, particularly for odontogenic keratocyst (OKC) and ameloblastoma (AM), with F1-score improvements ranging from 6.2% to 12.7%. For segmentation, the DSC was 87.2% and the ASSD was 1.359 mm. The model's average segmentation time was 40 ± 9.9 s, contrasting with 25 ± 7.2 min for OMSs. The proposed DL model accurately and efficiently classified and segmented five classes of jaw lesions using CBCT. In addition, it could assist doctors in improving classification accuracy and segmentation efficiency, particularly in distinguishing confusing lesions (eg, AM and OKC).

To evaluate the performance, agreement, and efficiency of a fully convolutional network (FCN) for liver lesion detection and segmentation at CT examinations in patients with colorectal liver metastases (CLMs). This retrospective study evaluated an automated method using an FCN that was trained, validated, and tested with 115, 15, and 26 contrast material-enhanced CT examinations containing 261, 22, and 105 lesions, respectively. Manual detection and segmentation by a radiologist was the reference standard. Performance of fully automated and user-corrected segmentations was compared with that of manual segmentations. The interuser agreement and interaction time of manual and user-corrected segmentations were assessed. Analyses included sensitivity and positive predictive value of detection, segmentation accuracy, Cohen κ, Bland-Altman analyses, and analysis of variance. In the test cohort, for lesion size smaller than 10 mm (n = 30), 10-20 mm (n = 35), and larger than 20 mm (n = 40), the detection sensitivity of the automated method was 10%, 71%, and 85%; positive predictive value was 25%, 83%, and 94%; Dice similarity coefficient was 0.14, 0.53, and 0.68; maximum symmetric surface distance was 5.2, 6.0, and 10.4 mm; and average symmetric surface distance was 2.7, 1.7, and 2.8 mm, respectively. For manual and user-corrected segmentation, κ values were 0.42 (95% confidence interval: 0.24, 0.63) and 0.52 (95% confidence interval: 0.36, 0.72); normalized interreader agreement for lesion volume was -0.10 ± 0.07 (95% confidence interval) and -0.10 ± 0.08; and mean interaction time was 7.7 minutes ± 2.4 (standard deviation) and 4.8 minutes ± 2.1 (P < .001), respectively. Automated detection and segmentation of CLM by using deep learning with convolutional neural networks, when manually corrected, improved efficiency but did not substantially change agreement on volumetric measurements.© RSNA, 2019Supplemental material is available for this article.