Development of a Micro-Macro Neural Network to recognize rollover movement

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Bone metastasis causes significant pain during trunk movements like rollovers.
• Existing methods for detecting rollover movements using electromyography (EMG) signals have limitations in speed and accuracy.
• Developing assistive systems is crucial for improving the quality of life for patients with cancer bone metastasis.

Purpose of the Study:

• To develop an effective rollover-support system for patients with cancer bone metastasis.
• To introduce a novel Micro-Macro Neural Network (MMNN) for earlier and more accurate recognition of rollover movements.
• To enhance the EMG signal processing for pain mitigation during trunk twists.

Main Methods:

• Development of a Micro-Macro Neural Network (MMNN) for EMG signal analysis.
• Comparison of MMNN performance against the traditional Time Delay Neural Network (TDNN).
• Utilizing EMG signals from the internal abdominal oblique muscle to detect trunk movements.

Main Results:

• MMNN achieved recognition 49 msec faster than TDNN (S.D. 45 msec).
• Pre-rollover recognition rate improved from 38% with TDNN to 86% with MMNN.
• MMNN reduced false recognitions to one-third of TDNN, demonstrating superior de-noising and historical data utilization.

Conclusions:

• The MMNN significantly outperforms TDNN in speed and accuracy for rollover movement recognition.
• MMNN's ability to effectively utilize past EMG data and its de-noising capabilities are key advantages.
• This research lays the groundwork for an effective rollover-support system for bone metastasis patients, aiming to reduce pain and improve mobility.