Faculty Bibliography

PURPOSE:To develop and evaluate a novel deep learning-based reconstruction framework called SANTIS (Sampling-Augmented Neural neTwork with Incoherent Structure) for efficient MR image reconstruction with improved robustness against sampling pattern discrepancy. METHODS:With a combination of data cycle-consistent adversarial network, end-to-end convolutional neural network mapping, and data fidelity enforcement for reconstructing undersampled MR data, SANTIS additionally utilizes a sampling-augmented training strategy by extensively varying undersampling patterns during training, so that the network is capable of learning various aliasing structures and thereby removing undersampling artifacts more effectively and robustly. The performance of SANTIS was demonstrated for accelerated knee imaging and liver imaging using a Cartesian trajectory and a golden-angle radial trajectory, respectively. Quantitative metrics were used to assess its performance against different references. The feasibility of SANTIS in reconstructing dynamic contrast-enhanced images was also demonstrated using transfer learning. RESULTS:Compared to conventional reconstruction that exploits image sparsity, SANTIS achieved consistently improved reconstruction performance (lower errors and greater image sharpness). Compared to standard learning-based methods without sampling augmentation (e.g., training with a fixed undersampling pattern), SANTIS provides comparable reconstruction performance, but significantly improved robustness, against sampling pattern discrepancy. SANTIS also achieved encouraging results for reconstructing liver images acquired at different contrast phases. CONCLUSION:By extensively varying undersampling patterns, the sampling-augmented training strategy in SANTIS can remove undersampling artifacts more robustly. The novel concept behind SANTIS can particularly be useful for improving the robustness of deep learning-based image reconstruction against discrepancy between training and inference, an important, but currently less explored, topic.

Purpose To investigate the diagnostic performance of three-dimensional (3D) MRI for depicting meniscal injuries of the knee by using surgery as the standard of reference. Materials and Methods A literature search was performed to identify original studies published between 1985 and 2017. Summary receiver operating characteristic curve and sensitivity analyses were performed to compare the diagnostic performance of 3D versus two-dimensional (2D) MRI for the assessment of knee meniscal injuries and to evaluate the impact of relevant covariates on the diagnostic performance for assessment of knee meniscal injuries. Results Of identified records, 31 studies (1743 3D knee MRI examinations) were included (23 studies also reported the results of 2D MRI). All studies before 2008 used gradient-echo (GRE) sequences, whereas all studies after 2011 used fast spin-echo (FSE) sequences. By comparing FSE and GRE sequences with 2D MRI, pooled estimate of sensitivity (90.0%; P = .2 and 90.1%; P = .2 vs 88.5%) and pooled estimate of specificity (91%; P = .3 and 89.8% vs 90.1%; P = .7) were comparable. The 3D FSE sequences demonstrated similar diagnostic performance as 3D GRE sequences, except for slightly improved sensitivity for depicting lateral meniscal injuries (FSE, 84.6%; GRE, 75%; P = .01). The specificity of 3D sequences improved when multiplanar reformatting was performed (P = .02). Conclusion Both three-dimensional (3D) fast spin-echo (FSE) and 3D gradient-echo (GRE) sequences had similar diagnostic performance as two-dimensional sequences, with slight superior sensitivity of 3D FSE sequences compared with 3D GRE sequences for depicting lateral meniscal injuries of the knee. © RSNA, 2018 Online supplemental material is available for this article.

BACKGROUND:Despite the advantages of 3D MRI in evaluation of cruciate ligament injuries, its use in clinical practice is still a matter of debate due to controversy regarding its diagnostic performance. PURPOSE/OBJECTIVE:To evaluate the diagnostic performance of 3D MRI for detecting cruciate ligament injuries, using surgery or arthroscopy as the reference standard. STUDY TYPE/METHODS:Meta-analysis. POPULATION/METHODS:Patients with knee pain. FIELD STRENGTH/SEQUENCE/UNASSIGNED:3D and 2D MRI. ASSESSMENT/RESULTS:Four databases were reviewed according to PRISMA guidelines. STATISTICAL TESTS/UNASSIGNED:Pooled values of sensitivity, specificity, and diagnostic odds ratio (DOR) were calculated using a random-effects model. To investigate the effect of relevant covariates on the diagnostic performance of 3D MRI, sensitivity analysis was performed using meta-regression to calculate relative DOR. RESULTS:Of 731 initially identified reports, 22 (1298 3D MRI examinations) met our criteria and were included. Pooled estimates of sensitivity and specificity for 3D sequences were 91.4% (95% confidence interval [CI]: 87.4-94.2%) and 96.1% (95% CI: 93.8-97.6%), respectively. Fourteen studies also reported the results of 2D MRI, with pooled sensitivity of 90.6% (95% CI: 84.1-94.6%) and specificity of 97.1% (95% CI: 94.7-98.4%), which were not significantly different from 3D sequences. 3D MRI sequences performed using 3T scanners had significantly higher DOR compared with 3D sequences performed on 1.5T or lower scanners (relative DOR: 6.04, P = 0.01). DATA CONCLUSION/UNASSIGNED:3D MRI is equivalent to 2D MRI in the diagnosis of cruciate ligament injuries. The use of 3T scanners improves the performance of 3D MRI for detecting cruciate ligament injuries. LEVEL OF EVIDENCE/METHODS:2 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;50:1545-1560.

PURPOSE:To describe and evaluate a new segmentation method using deep convolutional neural network (CNN), 3D fully connected conditional random field (CRF), and 3D simplex deformable modeling to improve the efficiency and accuracy of knee joint tissue segmentation. METHODS:A segmentation pipeline was built by combining a semantic segmentation CNN, 3D fully connected CRF, and 3D simplex deformable modeling. A convolutional encoder-decoder network was designed as the core of the segmentation method to perform high resolution pixel-wise multi-class tissue classification for 12 different joint structures. The 3D fully connected CRF was applied to regularize contextual relationship among voxels within the same tissue class and between different classes. The 3D simplex deformable modeling refined the output from 3D CRF to preserve the overall shape and maintain a desirable smooth surface for joint structures. The method was evaluated on 3D fast spin-echo (3D-FSE) MR image data sets. Quantitative morphological metrics were used to evaluate the accuracy and robustness of the method in comparison to the ground truth data. RESULTS:The proposed segmentation method provided good performance for segmenting all knee joint structures. There were 4 tissue types with high mean Dice coefficient above 0.9 including the femur, tibia, muscle, and other non-specified tissues. There were 7 tissue types with mean Dice coefficient between 0.8 and 0.9 including the femoral cartilage, tibial cartilage, patella, patellar cartilage, meniscus, quadriceps and patellar tendon, and infrapatellar fat pad. There was 1 tissue type with mean Dice coefficient between 0.7 and 0.8 for joint effusion and Baker's cyst. Most musculoskeletal tissues had a mean value of average symmetric surface distance below 1 mm. CONCLUSION:The combined CNN, 3D fully connected CRF, and 3D deformable modeling approach was well-suited for performing rapid and accurate comprehensive tissue segmentation of the knee joint. The deep learning-based segmentation method has promising potential applications in musculoskeletal imaging.

Purpose To develop and evaluate the feasibility of deep learning approaches for magnetic resonance (MR) imaging-based attenuation correction (AC) (termed deep MRAC) in brain positron emission tomography (PET)/MR imaging. Materials and Methods A PET/MR imaging AC pipeline was built by using a deep learning approach to generate pseudo computed tomographic (CT) scans from MR images. A deep convolutional auto-encoder network was trained to identify air, bone, and soft tissue in volumetric head MR images coregistered to CT data for training. A set of 30 retrospective three-dimensional T1-weighted head images was used to train the model, which was then evaluated in 10 patients by comparing the generated pseudo CT scan to an acquired CT scan. A prospective study was carried out for utilizing simultaneous PET/MR imaging for five subjects by using the proposed approach. Analysis of covariance and paired-sample t tests were used for statistical analysis to compare PET reconstruction error with deep MRAC and two existing MR imaging-based AC approaches with CT-based AC. Results Deep MRAC provides an accurate pseudo CT scan with a mean Dice coefficient of 0.971 ± 0.005 for air, 0.936 ± 0.011 for soft tissue, and 0.803 ± 0.021 for bone. Furthermore, deep MRAC provides good PET results, with average errors of less than 1% in most brain regions. Significantly lower PET reconstruction errors were realized with deep MRAC (-0.7% ± 1.1) compared with Dixon-based soft-tissue and air segmentation (-5.8% ± 3.1) and anatomic CT-based template registration (-4.8% ± 2.2). Conclusion The authors developed an automated approach that allows generation of discrete-valued pseudo CT scans (soft tissue, bone, and air) from a single high-spatial-resolution diagnostic-quality three-dimensional MR image and evaluated it in brain PET/MR imaging. This deep learning approach for MR imaging-based AC provided reduced PET reconstruction error relative to a CT-based standard within the brain compared with current MR imaging-based AC approaches. ^© RSNA, 2017 Online supplemental material is available for this article.

PURPOSE:To describe and evaluate a new fully automated musculoskeletal tissue segmentation method using deep convolutional neural network (CNN) and three-dimensional (3D) simplex deformable modeling to improve the accuracy and efficiency of cartilage and bone segmentation within the knee joint. METHODS:A fully automated segmentation pipeline was built by combining a semantic segmentation CNN and 3D simplex deformable modeling. A CNN technique called SegNet was applied as the core of the segmentation method to perform high resolution pixel-wise multi-class tissue classification. The 3D simplex deformable modeling refined the output from SegNet to preserve the overall shape and maintain a desirable smooth surface for musculoskeletal structure. The fully automated segmentation method was tested using a publicly available knee image data set to compare with currently used state-of-the-art segmentation methods. The fully automated method was also evaluated on two different data sets, which include morphological and quantitative MR images with different tissue contrasts. RESULTS:The proposed fully automated segmentation method provided good segmentation performance with segmentation accuracy superior to most of state-of-the-art methods in the publicly available knee image data set. The method also demonstrated versatile segmentation performance on both morphological and quantitative musculoskeletal MR images with different tissue contrasts and spatial resolutions. CONCLUSION:The study demonstrates that the combined CNN and 3D deformable modeling approach is useful for performing rapid and accurate cartilage and bone segmentation within the knee joint. The CNN has promising potential applications in musculoskeletal imaging. Magn Reson Med 79:2379-2391, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

BACKGROUND:Abnormal knee mechanics may contribute to early cartilage degeneration following anterior cruciate ligament reconstruction. Anterior cruciate ligament graft geometry has previously been linked to abnormal tibiofemoral kinematics, suggesting this parameter may be important in restoring normative cartilage loading. However, the relationship between graft geometry and cartilage contact is unknown. METHODS:Static MR images were collected and segmented for eighteen subjects to obtain bone, cartilage, and anterior cruciate ligament geometries for their reconstructed and contralateral knees. The footprint locations and orientation of the anterior cruciate ligament were calculated. Volumetric, dynamic MR imaging was also performed to measure tibiofemoral kinematics, cartilage contact location, and contact sliding velocity while subjects performed loaded knee flexion-extension. Multiple linear regression was used to determine the relationship between non-anatomic graft geometry and asymmetric knee mechanics. FINDINGS: = 0.54). INTERPRETATION:This study provides evidence that non-anatomic graft geometry is linked to asymmetric knee mechanics, suggesting that restoring native anterior cruciate ligament geometry may be important to mitigate the risk of early cartilage degeneration in these patients.

BACKGROUND:F-FDG) PET images. A deep convolutional encoder-decoder network was trained to identify tissue contrast in volumetric uncorrected PET images co-registered to CT data. A set of 100 retrospective 3D FDG PET head images was used to train the model. The model was evaluated in another 28 patients by comparing the generated pseudo-CT to the acquired CT using Dice coefficient and mean absolute error (MAE) and finally by comparing reconstructed PET images using the pseudo-CT and acquired CT for attenuation correction. Paired-sample t tests were used for statistical analysis to compare PET reconstruction error using deepAC with CT-based attenuation correction. RESULTS:F-FDG PET results with average errors of less than 1% in most brain regions. CONCLUSIONS:F-FDG non-attenuation-corrected (NAC) PET image and evaluated it in PET/CT brain imaging.

OBJECTIVE:The objective of our study was to use a T2 mapping sequence performed at 3 T to investigate changes in the composition and microstructure of the cartilage and menisci of the pediatric knee joint during maturation. MATERIALS AND METHODS:This retrospective study was performed of MRI examinations of 76 pediatric knees without internal derangement in 72 subjects (29 boys [mean age, 12.5 years] and 43 girls [mean age, 13.0 years]) who were evaluated with a sagittal T2 mapping sequence. T2 relaxation time values were quantitatively measured in eight cartilage subregions and in the medial and lateral menisci. Wilcoxon rank sum and Kruskal-Wallis tests were used to analyze the relationship between cartilage and meniscus T2 relaxation time values and sex and skeletal maturation, respectively. A multivariate linear regression model was used to investigate the independent association between cartilage T2 relaxation time values and age, weight, and body mass index (BMI [weight in kilograms divided by the square of height in meters]). RESULTS:There were no significant sex differences (p = 0.26-0.91) in T2 relaxation time values for cartilage or meniscus. T2 relaxation time values in each individual cartilage subregion significantly decreased (p < 0.001) with progressive maturation. T2 relaxation time values in the lateral meniscus significantly increased (p = 0.001) with maturation, whereas T2 relaxation time values in the medial meniscus did not significantly change (p = 0.82). There was a significant association (p < 0.001) between cartilage T2 relaxation time values and age independent of weight and BMI, but no significant association between cartilage T2 relaxation time values and weight (p = 0.06) and BMI (p = 0.20) independent of age. CONCLUSION:Cartilage T2 relaxation time values significantly decreased in all cartilage subregions and meniscus T2 relaxation time values significantly increased in the lateral meniscus during maturation. These changes in T2 relaxation time values reflect age-related changes in tissue composition and microstructure.

Purpose To determine whether a T2 mapping sequence could depict early changes in the composition and microstructure of cartilage overlying stable lesions of the medial femoral condyle in patients with juvenile osteochondritis dissecans (JOCD). Materials and Methods This retrospective study analyzed a sagittal T2 mapping sequence performed between September 1, 2015, and March 31, 2017, on 16 patients (10 boys and six girls; median age, 11.5 years) with 18 stable medial femoral condyle JOCD lesions and 18 age-, sex-, and skeletal maturation-matched control participants (11 boys and seven girls; median age, 11.5 years). Cartilage T2 values were quantitatively measured within regions of interest placed around the cartilage within and overlying the JOCD lesion in patients with JOCD and around the cartilage on the weight-bearing medial femoral condyle in patients with JOCD and controls. Wilcoxon signed rank and Wilcoxon rank sum tests were used to compare T2 values. Results T2 values were significantly higher (P < .001) for cartilage within the JOCD lesion than for cartilage overlying the JOCD lesion in patients with JOCD. However, there were no significant differences in T2 values between cartilage overlying the JOCD lesion and cartilage on the weight-bearing medial femoral condyle in patients with JOCD (P = .67) or in T2 values of the cartilage on the weight-bearing medial femoral condyle between patients with JOCD and controls (P = .30). Conclusion There were no significant quantifiable differences in T2 values of cartilage overlying stable JOCD lesions and normal cartilage on the medial femoral condyle, suggesting no substantial changes in cartilage composition and microstructure.