Abstract
Accurate preoperative templating in total hip arthroplasty (THA) improves the procedure's precision, shortens its duration, and reduces complications such as instability and dislocations, prosthesis loosening, and loss of bone stock, among others. The current standard method uses X-ray images to manually determine implant size; however, the emergence of artificial intelligence (AI) is gaining increasing interest in its use for preoperative templating. This systematic review and meta-analysis aims to assess the difference in accuracy between the standard and AI methods for preoperative templating in regards to implant size. A systematic search of electronic databases identified 36 papers related to AI in preoperative planning in THA. Of these, 9 studies met the inclusion criteria and cumulatively yielded 1049 patient for comparison. Statistical analysis was performed using OpenMetaAnalysis software, utilising a random effects model with reported results at 95% confidence intervals. Results showed the odds for exact size match using an AI-assisted method were 4.163 times higher than the standard manual method in predicting an acetabular cup component (OR=4.163, P<0.001) and 3.672 times higher in predicting a femoral stem component (OR=3.672, P<0.001). Moreover, operative time following the use of the AI-assisted method was 9.2 minutes less than operations following the surgeon-performed method (MD=-9.2, P=0.027), although this was reduced to 4.35 minutes when the source of heterogeneity was removed (MD=-4.35, P=0.025). A key limitation of this study is that all the papers identified in the literature and included in this meta-analysis originated from China, thus limiting generalisability to other healthcare systems and populations. It was concluded that AI-assisted preoperative templating is significantly more accurate than the standard templating method currently used in clinical practice in predicting implant size in total hip arthroplasty and it helps to reduce operating time, although a high level of evidence from different centres worldwide is still lacking to validate its use in clinical practice.