The identification of bone regions for the assistance of temporal bone surgery through analysis of the drill acoustics

This paper presents a method to identify the bone regions encountered during a temporal bone dissection. This should aid the otolaryngologist in the identification of anatomical landmarks whilst training and performing surgical procedures. The approach adopts the mel frequency cepstrum coefficients and the Mahalanobis distance metric to indentify the temporal bone regions from the acoustic signature of the surgical drill. Audio data from a mastoidectomy, that involved the dissection of the sigmoid sinus, thin tegmen and thick tegmen bone regions with diamond and cutting type burrs of sizes 6 mm, 3 mm and 1 mm were analysed. An identification accuracy of 97.72 ± 1.77% was achieved with the mel frequency cepstrum approach. This is a significant improvement over the previous approach that utilises the power spectral density and the correlation coefficient to identify the same bone regions only 90.59 ± 1.58% of the time. Three case studies were considered as a feasibility study for the implementation of this method into a smart embedded drill environment. With a priori knowledge of the burr type, an identification accuracy of 96.54 ± 4.12% was achieved. When the burr type was not considered this accuracy decreased to 94.2 ± 3.17%. An accuracy of 90.88 ± 7.53% was obtained when identifying both the tool type and the bone region from the drill sound.