Automated 3D thorax model generation using handheld video-footage.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Anatomy

Background:

• Standard Electrical Impedance Tomography (EIT) reconstruction models often use generic thorax dimensions and electrode placements.
• Discrepancies between standard models and individual patient anatomy or electrode positioning lead to inaccurate impedance distribution visualization.

Purpose of the Study:

• To develop and evaluate a method for generating patient-specific 3D thorax models and electrode locations using handheld video footage.
• To address inaccuracies in EIT visualization caused by non-patient-specific models.

Main Methods:

• A process was developed to capture patient chest and electrode data using smartphone video.
• Structure from motion techniques were employed to generate 3D models from extracted images.
• ArUco markers were used to precisely locate electrode positions within the 3D models.

Main Results:

• Automated 3D model reconstruction from handheld video or images was achieved.
• The system generates sparse point clouds and reconstructs surface meshes, providing relative electrode coordinates.
• Average mean distance error was 5.4 mm, with a mean standard deviation of 6.0 mm. Reconstruction time averaged 5:17 minutes.

Conclusions:

• Thorax and electrode model reconstruction using readily available devices at emergency sites is feasible.
• This work represents a significant step towards automated, patient-specific EIT reconstruction models derived from handheld device imagery.