Bone Age Assessment Using Artificial Intelligence in Korean Pediatric Population: A Comparison of Deep-Learning Models Trained With Healthy Chronological and Greulich-Pyle Ages as Labels

Objective@#To develop a deep-learning-based bone age prediction model optimized for Korean children and adolescents and evaluate its feasibility by comparing it with a Greulich-Pyle-based deep-learning model. @*Materials and Methods@#A convolutional neural network was trained to predict age according to the bone development shown on a hand radiograph (bone age) using 21036 hand radiographs of Korean children and adolescents without known bone development-affecting diseases/conditions obtained between 1998 and 2019 (median age [interquartile range {IQR}], 9 [7–12] years; male:female, 11794:9242) and their chronological ages as labels (Korean model). We constructed 2 separate external datasets consisting of Korean children and adolescents with healthy bone development (Institution 1: n = 343;median age [IQR], 10 [4–15] years; male: female, 183:160; Institution 2: n = 321; median age [IQR], 9 [5–14] years; male:female, 164:157) to test the model performance. The mean absolute error (MAE), root mean square error (RMSE), and proportions of bone age predictions within 6, 12, 18, and 24 months of the reference age (chronological age) were compared between the Korean model and a commercial model (VUNO Med-BoneAge version 1.1; VUNO) trained with Greulich-Pyle-based age as the label (GP-based model). @*Results@#Compared with the GP-based model, the Korean model showed a lower RMSE (11.2 vs. 13.8 months; P = 0.004) and MAE (8.2 vs. 10.5 months; P = 0.002), a higher proportion of bone age predictions within 18 months of chronological age (88.3% vs. 82.2%; P = 0.031) for Institution 1, and a lower MAE (9.5 vs. 11.0 months; P = 0.022) and higher proportion of bone age predictions within 6 months (44.5% vs. 36.4%; P = 0.044) for Institution 2. @*Conclusion@#The Korean model trained using the chronological ages of Korean children and adolescents without known bone development-affecting diseases/conditions as labels performed better in bone age assessment than the GP-based model in the Korean pediatric population. Further validation is required to confirm its accuracy.

Re-Assessment of Applicability of Greulich and Pyle-Based Bone Age to Korean Children Using Manual and Deep Learning-Based Automated Method

Purpose@#To evaluate the applicability of Greulich-Pyle (GP) standards to bone age (BA) assessment in healthy Korean children using manual and deep learning-based methods. @*Materials and Methods@#We collected 485 hand radiographs of healthy children aged 2–17 years (262 boys) between 2008 and 2017. Based on GP method, BA was assessed manually by two radiologists and automatically by two deep learning-based BA assessment (DLBAA), which estimated GP-assigned (original model) and optimal (modified model) BAs. Estimated BA was compared to chronological age (CA) using intraclass correlation (ICC), Bland-Altman analysis, linear regression, mean absolute error, and root mean square error. The proportion of children showing a difference >12 months between the estimated BA and CA was calculated. @*Results@#CA and all estimated BA showed excellent agreement (ICC ≥0.978, p12 months in 44.3%, 44.5%, 39.2%, and 36.1% for radiologist 1, radiologist 2, original, and modified DLBAA models, respectively. @*Conclusion@#Contemporary healthy Korean children showed different rates of skeletal development than GP standard-BA, and systemic bias should be considered when determining children’s skeletal maturation.

Diagnosis of Nerve Root Compromise of the Lumbar Spine: Evaluation of the Performance of Three-dimensional Isotropic T2-weighted Turbo Spin-Echo SPACE Sequence at 3T

OBJECTIVE: To explore the performance of three-dimensional (3D) isotropic T2-weighted turbo spin-echo (TSE) sampling perfection with application optimized contrasts using different flip angle evolution (SPACE) sequence on a 3T system, for the evaluation of nerve root compromise by disc herniation or stenosis from central to extraforaminal location of the lumbar spine, when used alone or in combination with conventional two-dimensional (2D) TSE sequence. MATERIALS AND METHODS: Thirty-seven patients who had undergone 3T spine MRI including 2D and 3D sequences, and had subsequent spine surgery for nerve root compromise at a total of 39 nerve levels, were analyzed. A total of 78 nerve roots (48 symptomatic and 30 asymptomatic sites) were graded (0 to 3) using different MRI sets of 2D, 3D (axial plus sagittal), 3D (all planes), and combination of 2D and 3D sequences, with respect to the nerve root compromise caused by posterior disc herniations, lateral recess stenoses, neural foraminal stenoses, or extraforaminal disc herniations; grading was done independently by two readers. Diagnostic performance was compared between different imaging sets using the receiver operating characteristics (ROC) curve analysis. RESULTS: There were no statistically significant differences (p = 0.203 to > 0.999) in the ROC curve area between the imaging sets for both readers 1 and 2, except for combined 2D and 3D (0.843) vs. 2D (0.802) for reader 1 (p = 0.035), and combined 2D and 3D (0.820) vs. 3D including all planes (0.765) for reader 2 (p = 0.049). CONCLUSION: The performance of 3D isotropic T2-weighted TSE sequence of the lumbar spine, whether axial plus sagittal images, or all planes of images, was not significantly different from that of 2D TSE sequences, for the evaluation of nerve root compromise of the lumbar spine. Combining 2D and 3D might possibly improve the diagnostic accuracy compared with either one.

Subtraction MR Venography Acquired from Time-Resolved Contrast-Enhanced MR Angiography: Comparison with Phase-Contrast MR Venography and Single-Phase Contrast-Enhanced MR Venography

OBJECTIVE: To evaluate the image characteristics of subtraction magnetic resonance venography (SMRV) from time-resolved contrast-enhanced MR angiography (TRMRA) compared with phase-contrast MR venography (PCMRV) and single-phase contrast-enhanced MR venography (CEMRV). MATERIALS AND METHODS: Twenty-one patients who underwent brain MR venography (MRV) using standard protocols (PCMRV, CEMRV, and TRMRA) were included. SMRV was made by subtracting the arterial phase data from the venous phase data in TRMRA. Co-registration and subtraction of the two volume data was done using commercially available software. Image quality and the degree of arterial contamination of the three MRVs were compared. In the three MRVs, 19 pre-defined venous structures (14 dural sinuses and 5 cerebral veins) were evaluated. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the three MRVs were also compared. RESULTS: Single-phase contrast-enhanced MR venography showed better image quality (median score 4 in both reviewers) than did the other two MRVs (p < 0.001), whereas SMRV (median score 3 in both reviewers) and PCMRV (median score 3 in both reviewers) had similar image quality (p ≥ 0.951). SMRV (median score 0 in both reviewers) suppressed arterial signal better than did the other MRVs (median score 1 in CEMRV, median score 2 in PCMRV, both reviewers) (p < 0.001). The dural sinus score of SMRV (median and interquartile range [IQR] 48, 43-50 for reviewer 1, 47, 43-49 for reviewer 2) was significantly higher than for PCMRV (median and IQR 31, 25-34 for reviewer 1, 30, 23-32 for reviewer 2) (p < 0.01) and did not differ from that of CEMRV (median and IQR 50, 47-52 for reviewer 1, 49, 45-51 for reviewer 2) (p = 0.146 in reviewer 1 and 0.123 in reviewer 2). The SNR and CNR of SMRV (median and IQR 104.5, 83.1-121.2 and 104.1, 74.9-120.5, respectively) were between those of CEMRV (median and IQR 150.3, 111-182.6 and 148.4, 108-178.2) and PCMRV (median and IQR 59.4, 49.2-74.9 and 53.6, 43.8-69.2). CONCLUSION: Subtraction magnetic resonance venography is a promising MRV method, with acceptable image quality and good arterial suppression.