Acoustic evaluation by artificial intelligence for initial press-fit cementless cup fixation in total hip arthroplasty

Abstract

Aims:

This study aimed to assess the evaluation accuracy of an artificial intelligence (AI) algorithm using the hammering sound to assess initial press-fit fixation in cementless total hip arthroplasty (THA). The acoustic features of the hammering sounds during a press-fit cup fixation were recorded and analysed.

Methods:

81 hips of 79 patients undergoing primary THA using the Trident HA acetabular system (Stryker, Kalamazoo, MI, USA) for treatment of osteoarthritis and osteonecrosis of the femoral head were included. After the fast Fourier transform analysis was performed on the hammering sound, 24 sound pressure (SP) features across 23 frequency bands were extracted and applied to binary classification. Support vector machine algorithms were used to classify the data using 2 models: Model A - SP only; Model B - SP + patient characteristics.

Results:

Among the 81 hips in 79 patients, the hammering sound of 348 impacts of 70 successful press-fit of 1^st attempts in 70 hips and 80 impacts of 16 failed attempts in 11hips were analysed. The area under the curve for the test data was 0.9554 in Model A and 1.0 in Model B. The accuracy (accuracy/sensitivity/specificity/positive predictive value/negative predictive value) of each model in the test data was as follows: Model A, 0.849/1.000/0.188/0.843/1.000 and Model B, 0.965/1.000/0.813/0.959/1.000.

Conclusions:

The accuracy of an AI algorithm using hammering sounds to judge the success of initial press-fit fixation in cementless cup in THA was relatively high. Our results imply the possibility of a practical application for such a system to assist surgeons during the procedure.