Artificial intelligence in commercial fracture detection products: a systematic review and meta-analysis of diagnostic test accuracy.

Conventional radiography (CR) is primarily utilized for fracture diagnosis. Artificial intelligence (AI) for CR is a rapidly growing field aimed at enhancing efficiency and increasing diagnostic accuracy. However, the diagnostic performance of commercially available AI fracture detection solutions (CAAI-FDS) for CR in various anatomical regions, their synergy with human assessment, as well as the influence of industry funding on reported accuracy are unknown. Peer-reviewed diagnostic test accuracy (DTA) studies were identified through a systematic review on Pubmed and Embase. Diagnostic performance measures were extracted especially for different subgroups such as product, type of rater (stand-alone AI, human unaided, human aided), funding, and anatomical region. Pooled measures were obtained with a bivariate random effects model. The impact of rater was evaluated with comparative meta-analysis. Seventeen DTA studies of seven CAAI-FDS analyzing 38,978 x-rays with 8,150 fractures were included. Stand-alone AI studies (n = 15) evaluated five CAAI-FDS; four with good sensitivities (> 90%) and moderate specificities (80-90%) and one with very poor sensitivity (< 60%) and excellent specificity (> 95%). Pooled sensitivities were good to excellent, and specificities were moderate to good in all anatomical regions (n = 7) apart from ribs (n = 4; poor sensitivity / moderate specificity) and spine (n = 4; excellent sensitivity / poor specificity). Funded studies (n = 4) had higher sensitivity (+ 5%) and lower specificity (-4%) than non-funded studies (n = 11). Sensitivity did not differ significantly between stand-alone AI and human AI aided ratings (p = 0.316) but specificity was significantly higher the latter group (p < 0.001). Sensitivity was significant lower in human unaided compared to human AI aided respectively stand-alone AI ratings (both p ≤ 0.001); specificity was higher in human unaided ratings compared to stand-alone AI (p < 0.001) and showed no significant differences AI aided ratings (p = 0.316). The study demonstrates good diagnostic accuracy across most CAAI-FDS and anatomical regions, with the highest performance achieved when used in conjunction with human assessment. Diagnostic accuracy appears lower for spine and rib fractures. The impact of industry funding on reported performance is small.

Development of the humeral head offset index for control of humeral torsion.

BACKGROUND: Control of humeral torsion can present a challenge, especially intraoperatively during closed reduction and fixation of humeral shaft fractures or 2-part surgical neck fractures of the proximal humerus. The objective of this study is to develop and validate an indirect method for the assessment of humeral torsion using an index that is linearly correlated with rotational arm position and can be derived from only a single plain radiographic image of the proximal humerus. METHODS: The Humeral Head Offset Index (HHOI) is calculated as the ratio of the medial and lateral offset of the humeral head measured from the outer cortices of the shaft on a plain radiographic or fluoroscopic image. The relationship of HHOI with humeral torsion was first verified on a sawbone model with radiopaque characteristics under fluoroscopic control. Different degrees of retroversion were simulated through manual rotation of the humerus with a digital protractor in 5° increments until 40° internally rotated and then in 5° increments until 40° externally rotated from the neutral position. The same procedure was subsequently performed digitally on Digitally Reconstructed Radiographs (DRRs) from computed tomography (CT) dataset of the sawbone. Next, the HHOI index was applied to eight randomly selected patients with total humerus CT using the same method. Spearman's rho was calculated for the bivariate analysis of correlation between the simulated degree of retroversion and the HHOI. Strength of correlation was classified according to Koo and Li. Interrater and intrarater reliability of three blinded observers with repetition of measurement after three months were analyzed by assessing the intraclass correlation coefficient (ICC). RESULTS: Both in the sawbone model and in DRRs, we demonstrated a high to very high significant linear correlation between simulated retroversion and the HHOI. ICC values demonstrated excellent interrater reliability and excellent intrarater reliability for measurement of the HHOI. CONCLUSIONS: The HHOI is a new, simple, reliable index that has a linear relationship to the rotation of the humerus and can therefore allow an indirect control of humeral torsion in comparison to the contralateral side.

Applications and accuracy of 3D-printed surgical guides in traumatology and orthopaedic surgery: A systematic review and meta-analysis.

Background: Patient-Specific Surgical Guides (PSSGs) are advocated for reducing radiation exposure, operation time and enhancing precision in surgery. However, existing accuracy assessments are limited to specific surgeries, leaving uncertainties about variations in accuracy across different anatomical sites, three-dimensional (3D) printing technologies and manufacturers (traditional vs. printed at the point of care). This study aimed to evaluate PSSGs accuracy in traumatology and orthopaedic surgery, considering anatomical regions, printing methods and manufacturers. Methods: A systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. Studies were eligible if they (1) assessed the accuracy of PSSGs by comparing preoperative planning and postoperative results in at least two different planes (2) used either computer tomography or magnetic resonance imaging (3) covered the field of orthopaedic surgery or traumatology and (4) were available in English or German language. The 'Quality Assessment Tool for Quantitative Studies' was used for methodological quality assessment. Descriptive statistics, including mean, standard deviation, and ranges, are presented. A random effects meta-analysis was performed to determine the pooled mean absolute deviation between preoperative plan and postoperative result for each anatomic region (shoulder, hip, spine, and knee). Results: Of 4212 initially eligible studies, 33 were included in the final analysis (8 for shoulder, 5 for hip, 5 for spine, 14 for knee and 1 for trauma). Pooled mean deviation (95% confidence interval) for total knee arthroplasty (TKA), total shoulder arthroplasty (TSA), total hip arthroplasty (THA) and spine surgery (pedicle screw placement during spondylodesis) were 1.82° (1.48, 2.15), 2.52° (1.9, 3.13), 3.49° (3.04, 3.93) and 2.67° (1.64, 3.69), respectively. Accuracy varied between TKA and THA and between TKA and TSA. Conclusion: Accuracy of PSSGs depends on the type of surgery but averages around 2-3° deviation from the plan. The use of PSSGs might be considered for selected complex cases. Level of Evidence: Level 3 (meta-analysis including Level 3 studies).