Exploring LSTM Networks for Vibration-Based Bone Layer Classification

 
Introduction

Bone layer classification is a crucial task in medical procedures involving drilling, such as orthopedic surgeries and dental implants. Accurate identification of bone layers—Periosteum, First Cortical, and Spongy—is essential to ensure precision and safety during these procedures. In this blog post, we explore the use of Long Short-Term Memory (LSTM) networks, a type of deep learning model particularly well-suited for sequential data, to classify bone layers based on vibration data collected during drilling operations.

Dataset

The dataset used in this portfolio is provided by Caesarendra et al. (2024) and is detailed in their research paper titled "Bone Drilling Vibration Signal Classification Using Convolutional Neural Network to Determine Bone Layers" (10.1007/978-981-97-1463-6_40). The dataset consists of vibration signals captured during the bone drilling process, with labels indicating the bone layer.

Understanding the Problem

During drilling, different bone layers exhibit unique vibration signatures. By analyzing these vibrations, we can potentially classify the bone layers. The dataset consists of the following features:

1. Layer: The type of bone layer (Periosteum, First Cortical, Spongy)
2. Time: Time of data recording
3. X-axis Vibration: Vibration data along the X-axis
4. Y-axis Vibration: Vibration data along the Y-axis
5. Z-axis Vibration: Vibration data along the Z-axis

Code:  Google Colab