Background and purpose - We aimed to evaluate the ability of artificial intelligence (a deep learning algorithm) to detect and classify proximal humerus fractures using plain anteroposterior shoulder radiographs. Patients and methods - 1,891 images (1 image per person) of normal shoulders (n = 515) and 4 proximal humerus fracture types (greater tuberosity, 346; surgical neck, 514; 3-part, 269; 4-part, 247) classified by 3 specialists were evaluated. We trained a deep convolutional neural network (CNN) after augmentation of a training dataset. The ability of the CNN, as measured by top-1 accuracy, area under receiver operating characteristics curve (AUC), sensitivity/specificity, and Youden index, in comparison with humans (28 general physicians, 11 general orthopedists, and 19 orthopedists specialized in the shoulder) to detect and classify proximal humerus fractures was evaluated. Results - The CNN showed a high performance of 96% top-1 accuracy, 1.00 AUC, 0.99/0.97 sensitivity/specificity, and 0.97 Youden index for distinguishing normal shoulders from proximal humerus fractures. In addition, the CNN showed promising results with 65-86% top-1 accuracy, 0.90-0.98 AUC, 0.88/0.83-0.97/0.94 sensitivity/specificity, and 0.71-0.90 Youden index for classifying fracture type. When compared with the human groups, the CNN showed superior performance to that of general physicians and orthopedists, similar performance to orthopedists specialized in the shoulder, and the superior performance of the CNN was more marked in complex 3- and 4-part fractures. Interpretation - The use of artificial intelligence can accurately detect and classify proximal humerus fractures on plain shoulder AP radiographs. Further studies are necessary to determine the feasibility of applying artificial intelligence in the clinic and whether its use could improve care and outcomes compared with current orthopedic assessments.

Since its development, deep learning has been quickly incorporated into the field of medicine and has had a profound impact. Since 2017, many studies applying deep learning-based diagnostics in the field of orthopedics have demonstrated outstanding performance. However, most published papers have focused on disease detection or classification, leaving some unsatisfactory reports in areas such as segmentation and prediction. This review introduces research published in the field of orthopedics classified according to disease from the perspective of orthopedic surgeons, and areas of future research are discussed. This paper provides orthopedic surgeons with an overall understanding of artificial intelligence-based image analysis and the information that medical data should be treated with low prejudice, providing developers and researchers with insight into the real-world context in which clinicians are embracing medical artificial intelligence.

Meniscal extrusion prevention would be important for restoring normal knee kinematics, even though the effect of graft extrusion after meniscal allograft transplantation (MAT) has not been clearly identified. When compared with standard rehabilitation protocols, delayed rehabilitation after lateral MAT could reduce graft extrusion. Cohort study; Level of evidence, 3. A total of 53 patients who underwent lateral MAT with the keyhole technique from June 2011 to March 2015 were included. Graft extrusion and lateral compartment articular cartilage loss on magnetic resonance imaging, joint space width (JSW) on weightbearing radiographs, and clinical outcomes (including subjective scores and functional tests) of 25 patients who underwent standard rehabilitation (group S) after lateral MAT were compared with those of 28 patients who underwent delayed rehabilitation (group D), which included immobilization during the first 3 weeks and use of unloading braces for 9 weeks. Mean ± SD follow-up times for groups S and D were 25.7 ± 6.4 and 24.5 ± 7.1 months, respectively. Although the postoperative clinical outcomes did not differ between the groups, graft extrusion (3.2 ± 1.5 mm vs 1.8 ± 1.6 mm) and relative percentage of extrusion (32.2% ± 14.6% vs 17.3% ± 16.6%) were greater in group S than in group D (P < .001 for both). The percentage of patients with graft extrusion (≥3 mm) was 52% in group S and 21.4% in group D. JSWs on Rosenberg view decreased by 0.1 mm in group S but increased by 0.3 mm in group D (P = .035). Kellgren-Lawrence grade and modified Outerbridge grade progressed by 44% and 44% among patients in group S and by 17.9% and 21.4% among patients in group D, respectively. There were significant correlations between the coronal graft extrusion and postoperative JSWs on full extension (–0.452) and Rosenberg (–0.410) views, Kellgren-Lawrence grade (0.727), and modified Outerbridge grade (0.732) on magnetic resonance imaging (P < .001 for all). Compared with standard rehabilitation, delayed rehabilitation showed less coronal graft extrusion and joint space narrowing on weightbearing and reduced the progression of arthrosis, although the rehabilitation protocols showed no differences in clinical outcomes.

The purpose of this study was to compare the clinical and radiological results of 2 different tibial fixations performed using bioabsorbable screws with added hydroxyapatite (HA) and pure poly-L-lactic acid (PLLA) screws in anterior cruciate ligament (ACL) reconstruction. A total of 394 patients who underwent arthroscopic ACL reconstruction between March 2009 and June 2012 were retrospectively reviewed. Of those, 172 patients who took the radiological and clinical evaluations at more than 2 years after surgery were enrolled and divided into 2 groups: PLLA group (n = 86) and PLLA-HA group (n = 86). Both groups were assessed by means of the Lysholm score, International Knee Documentation Committee (IKDC) subjective knee score, and Tegner activity score. Stability was evaluated using the KT-2000 arthrometer. Magnetic resonance imaging was performed to evaluate tibial tunnel widening, screw resorption, osteoingeration, and foreign body reactions. The PLLA-HA group showed significant reduction in the extent of tibial tunnel widening and foreign body reactions and significant increase in screw resorption compared to the pure PLLA group (p < 0.001 for both). In contrast, postoperative Lysholm score, Tegner activity score, IKDC score, and side-to-side difference on the KT-2000 arthrometer showed no significant differences between groups (p = 0.478, p = 0.906, p = 0.362, and p = 0.078, respectively). The PLLA group showed more significant widening in the proximal tibial tunnel than the PLLA-HA group (p = 0.001). In the correlation analysis, proximal tibial tunnel widening revealed a positive correlation with knee laxity (r = 0.866) and a negative correlation with Lysholm score (r = -0.753) (p < 0.01 for both). The HA added PLLA screws would be advantageous for tibial graft fixation by reducing tibial tunnel widening, improving osteointegration, and lowering foreign body reactions. Even though no clinically significant differences were noted between the pure PLLA group and PLLA-HA group, widening of the proximal area of the tibial tunnel showed a tendency to increase knee laxity measured using the KT-2000 arthrometer.

There is considerable debate on the recovery of rotator cuff muscle atrophy after rotator cuff repair. To evaluate the serial changes in supraspinatus muscle volume after rotator cuff repair by using semiautomatic segmentation software and to determine the relationship with functional outcomes. Case series; Level of evidence, 4. Seventy-four patients (mean age, 62.8 ± 8.8 years) who underwent arthroscopic rotator cuff repair and obtained 3 consecutive (preoperatively, immediately postoperatively, and later postoperatively [≥1 year postoperatively]) magnetic resonance imaging (MRI) scans having complete Y-views were included. We generated a 3-dimensional (3D) reconstructed model of the supraspinatus muscle by using in-house semiautomatic segmentation software (ITK-SNAP) and calculated both the 2-dimensional (2D) cross-sectional area and 3D volume of the muscle in 3 different views (Y-view, 1 cm medial to the Y-view [Y+1 view], and 2 cm medial to the Y-view [Y+2 view]) at the 3 time points. The area and volume changes at each time point were evaluated according to repair integrity. Later postoperative volumes were compared with immediately postoperative volumes, and their relationship with various clinical factors and the effect of higher volume increases on range of motion, muscle power, and visual analog scale pain and American Shoulder and Elbow Surgeons scores were evaluated. The interrater reliabilities were excellent for all measurements. Areas and volumes increased immediately postoperatively as compared with preoperatively; however, only volumes on the Y+1 view and Y+2 view significantly increased later postoperatively as compared with immediately postoperatively (P < .05). There were 9 patients with healing failure, and area and volume changes were significantly less later postoperatively compared with immediately postoperatively at all measurement points in these patients (P < .05). After omitting the patients with healing failure, volume increases later postoperatively became more prominent (P < .05) in the order of the Y+2 view, Y+1 view, and Y-view. Volume increases were higher in patients who healed successfully with larger tears (P = .040). Higher volume increases were associated only with an increase in abduction power (P = .029) and not with other outcomes. The supraspinatus muscle volume increased immediately postoperatively and continuously for at least 1 year after surgery. The increase was evident in patients who had larger tears and healed successfully and when measured toward the more medial portion of the supraspinatus muscle. The volume increases were associated with an increase in shoulder abduction power.