No abstract

Objectives To evaluate the performance of a deep convolutional neural network (DCNN) in detecting and classifying distal radius fractures, metal, and cast on radiographs using labels based on radiology reports. The secondary aim was to evaluate the effect of the training set size on the algorithm’s performance. Methods A total of 15,775 frontal and lateral radiographs, corresponding radiology reports, and a ResNet18 DCNN were used. Fracture detection and classification models were developed per view and merged. Incrementally sized subsets served to evaluate effects of the training set size. Two musculoskeletal radiologists set the standard of reference on radiographs (test set A). A subset (B) was rated by three radiology residents. For a per-study-based comparison with the radiology residents, the results of the best models were merged. Statistics used were ROC and AUC, Youden’s J statistic (J), and Spearman’s correlation coefficient (ρ). Results The models’ AUC/J on (A) for metal and cast were 0.99/0.98 and 1.0/1.0. The models’ and residents’ AUC/J on (B) were similar on fracture (0.98/0.91; 0.98/0.92) and multiple fragments (0.85/0.58; 0.91/0.70). Training set size and AUC correlated on metal (ρ = 0.740), cast (ρ = 0.722), fracture (frontal ρ = 0.947, lateral ρ = 0.946), multiple fragments (frontal ρ = 0.856), and fragment displacement (frontal ρ = 0.595). Conclusions The models trained on a DCNN with report-based labels to detect distal radius fractures on radiographs are suitable to aid as a secondary reading tool; models for fracture classification are not ready for clinical use. Bigger training sets lead to better models in all categories except joint affection. Key Points • Detection of metal and cast on radiographs is excellent using AI and labels extracted from radiology reports. • Automatic detection of distal radius fractures on radiographs is feasible and the performance approximates radiology residents. • Automatic classification of the type of distal radius fracture varies in accuracy and is inferior for joint involvement and fragment displacement.

Objectives Supine lumbar spine examinations underestimate body weight effects on neuroforaminal size. Therefore, our purpose was to evaluate size changes of the lumbar neuroforamina using supine and upright 3D tomography and to initially assess image quality compared with computed tomography (CT). Methods The lumbar spines were prospectively scanned in 48 patients in upright (3D tomographic twin robotic X-ray) and supine (30 with 3D tomography, 18 with CT) position. Cross-sectional area (CSA), cranio-caudal (CC), and ventro-dorsal (VD) diameters of foramina were measured by two readers and additionally graded in relation to the intervertebral disc height. Visibility of bone/soft tissue structures and image quality were assessed independently on a 5-point Likert scale for the 18 patients scanned with both modalities. Descriptive statistics, Wilcoxon’s signed-rank test (p < 0.05), and interreader reliability were calculated. Results Neuroforaminal size significantly decreased at all levels for both readers from the supine (normal intervertebral disc height; CSA 1.25 ± 0.32 cm2; CC 1.84 ± 0.24 cm2; VD 0.88 ± 0.16 cm2) to upright position (CSA 1.12 ± 0.34 cm2; CC 1.78 ± 0.24 cm2; VD 0.83 ± 0.16 cm2; each p < 0.001). Decrease in intervertebral disc height correlated with decrease in foraminal size (supine: CSA 0.88 ± 0.34 cm2; CC 1.39 ± 0.33 cm2; VD 0.87 ± 0.26 cm2; upright: CSA 0.83 ± 0.37 cm2, p = 0.010; CC 1.32 ± 0.33 cm2, p = 0.015; VD 0.80 ± 0.21 cm2, p = 0.021). Interreader reliability for area was fair to excellent (0.51–0.89) with a wide range for cranio-caudal (0.32–0.74) and ventro-dorsal (0.03–0.70) distances. Image quality was superior for CT compared with that for 3D tomography (p < 0.001; κ, CT = 0.66–0.92/3D tomography = 0.51–1.00). Conclusions The size of the lumbar foramina is smaller in the upright weight-bearing position compared with that in the supine position. Image quality, especially nerve root delineation, is inferior using 3D tomography compared to CT. Key Points • Weight-bearing examination demonstrates a decrease of the neuroforaminal size. • Patients with higher decrease in intervertebral disc showed a narrower foraminal size. • Image quality is superior with CT compared to 3D tomographic twin robotic X-ray at the lumbar spine.