Automatic algorithm for the characterization of sweat ducts in a three-dimensional fingerprint

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Machine Learning

Background:

• Accurate characterization of sweat ducts and pores is crucial for various biomedical applications.
• Existing methods for analyzing sweat gland structures face limitations in depth-dependent accuracy.

Purpose of the Study:

• To develop an automated algorithm for quantitative analysis of sweat ducts and pores.
• To create a novel hybrid fingerprint for identification and encryption using sweat pore and internal fingerprint data.

Main Methods:

• Utilized a convolutional neural network (CNN) with U-net architecture for image analysis.
• Employed ellipsoidal and elliptical approximations for 3D sweat ducts and en face sweat pores.
• Developed projection-based extraction for sweat pores to mitigate depth-dependent diameter variations.
• Superposed projection-based sweat pore images with Maximum Intensity Projection (MIP)-based internal fingerprints.

Main Results:

• Demonstrated that ellipsoid length and diameter can quantitatively describe sweat ducts.
• Showed potential for estimating resonance frequencies in millimeter (mm) wave and terahertz (THz) wave applications.
• Successfully extracted projection-based sweat pores, overcoming depth-related inaccuracies.
• Constructed a hybrid internal fingerprint by combining sweat pore and internal fingerprint data.

Conclusions:

• The developed CNN-based algorithm provides a robust method for analyzing sweat duct morphology.
• The quantitative descriptors of sweat ducts have implications for mmWave and THz wave applications.
• The novel hybrid fingerprint shows promise for advanced identification recognition and information encryption.