Mechanical Imaging of Soft Tissues With Miniature Climbing Robots

Area of Science:

• Biomechanics
• Medical Diagnostics
• Robotics

Background:

• Quantitative mechanical property mapping of skin and soft tissues is crucial for biomechanics and disease diagnostics.
• Current manual methods lack quantitative data and are subject to practitioner variability.
• Tactile sensors offer potential for increased sensitivity in mechanical measurements.

Purpose of the Study:

• To develop a noninvasive method for testing soft tissue mechanical properties using skin-crawling robots.
• To train a convolutional neural network for classifying lump size and depth based on robotic sensor data.
• To enhance systematic and repeatable data collection for tissue characterization.

Main Methods:

• Utilized previously developed skin-crawling robots equipped with custom cutometers or indenters.
• Collected mechanical property data from simulated tissue with embedded lumps.
• Trained a convolutional neural network (CNN) to classify lump size and depth.

Main Results:

• The CNN achieved high classification accuracy: 98.8% for cutometer and 99.6% for indenter for lump size.
• Accurate classification of lump size (0-10 mm diameter) and depth (1-5 mm) in simulated tissue.
• Demonstrated feasibility through a limited evaluation on human forearm skin.

Conclusions:

• The proposed robotic system provides a systematic and repeatable method for noninvasive soft tissue mechanical property assessment.
• The CNN-based classification demonstrates significant potential for accurate lump detection and characterization.
• Future work aims to improve noninvasive tissue testing capabilities for enhanced sensitivity and data collection.