A Machine Learning Approach to Discriminate Between Soft and Hard Bone Tissues Using Drilling Sounds

4 pages•Published: June 27, 2017

Vahid Zakeri, Francesco Fabri, Masashi Karasawa and Antony J. Hodgson

Abstract

Bone drilling is conducted in many surgical disciplines such as orthopedics, maxillofacial, and spine surgery. Most of these procedures involve drilling of different bone materials including hard (cortical) and soft (cancellous) tissues. Identifying these tissues is essential for surgeons to minimize damage to underlying nerves and vessels.
The sound signal generated during drilling is a valuable source of information that could potentially be employed. Such sounds can be captured readily and easily through non-contact sensors. Therefore, our goal in this preliminary study is to investigate whether drilling sounds can enable us to distinguish between cortical and cancellous tissues.
A bovine tibial bone was drilled, and the cortical and cancellous drilling sounds were captured. Each sound record was divided into small windows with a length of 50 ms and a 50% overlap. The window length was selected small, because our intended longer-term application is to provide the surgeon with near-real-time feedback. Short time Fourier Transform (STFT) coefficients were extracted from each window and were averaged accordingly to obtain p features. A support vector machine (SVM) algorithm was used for classification, and its accuracy was evaluated for different number of features (p). Two training/testing scenarios were considered, atlas (ATL) and leave-one-out (LOO).
The total accuracies for ATL and LOO were 100% and 93.8% respectively obtained for p=128. Our study on a single specimen demonstrated that it is possible to discriminate between cortical and cancellous bones based on relatively short 50 ms windows of drilling sounds.

Keyphrases: bone drilling, cancellous, cortical, drilling sound, machine learning, tissue identification

In: Klaus Radermacher and Ferdinando Rodriguez Y Baena (editors). CAOS 2017. 17th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery, vol 1, pages 276-279.

Links:	https://easychair.org/publications/paper/1s
	https://doi.org/10.29007/q1s2

BibTeX entry

@inproceedings{CAOS2017:Machine_Learning_Approach_Discriminate,
  author    = {Vahid Zakeri and Francesco Fabri and Masashi Karasawa and Antony J. Hodgson},
  title     = {A Machine Learning Approach to Discriminate Between Soft and Hard Bone Tissues Using Drilling Sounds},
  booktitle = {CAOS 2017. 17th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery},
  editor    = {Klaus Radermacher and Ferdinando Rodriguez Y Baena},
  series    = {EPiC Series in Health Sciences},
  volume    = {1},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2398-5305},
  url       = {/publications/paper/1s},
  doi       = {10.29007/q1s2},
  pages     = {276-279},
  year      = {2017}}

Download PDF Open PDF in browser