Exploratory analysis and framework for tissue classification based on vibroacoustic signals from needle-tissue...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Signal Processing

Background:

• Accurate surgical needle localization is critical for medical procedures like biopsies and injections.
• Current imaging devices (MRI, CT, US) face artifacts, hindering precise needle tip identification.
• A novel needle guidance technique is necessary to overcome existing imaging limitations.

Purpose of the Study:

• To investigate the potential of vibroacoustic signals for surgical needle localization.
• To develop and evaluate deep learning models for analyzing needle-induced signals.
• To establish a foundation for advanced needle guidance systems.

Main Methods:

• Generating vibroacoustic signals by moving a needle through a specialized phantom with animal tissue.
• Preprocessing the acquired vibroacoustic data.
• Converting data into Mel and continuous wavelet transform spectrogram representations.
• Utilizing deep learning models (NeedleNet, ResNet-34) for signal analysis.

Main Results:

• Demonstrated the feasibility of using vibroacoustic signals for needle localization.
• Successfully applied deep learning models to spectrograms of needle-induced signals.
• Identified spectrogram representations and deep learning architectures for further research.

Conclusions:

• Vibroacoustic signal analysis holds significant potential for enhancing surgical needle guidance.
• Deep learning approaches show promise in accurately interpreting these signals.
• Further research is warranted to optimize this technique for clinical application.