Deep learning model for ultrafast multifrequency optical property extractions for spatial frequency domain imaging.

Area of Science:

• Biomedical Imaging
• Optical Physics
• Computational Biology

Background:

• Spatial Frequency Domain Imaging (SFDI) is a label-free technique for mapping tissue optical properties.
• Traditional SFDI methods often use limited spatial frequencies, impacting accuracy and speed.
• Advanced multi-frequency SFDI offers improved accuracy but is computationally intensive.

Purpose of the Study:

• To develop a rapid and accurate method for solving the multi-frequency SFDI inverse problem.
• To overcome the computational bottleneck of existing multi-frequency SFDI algorithms.
• To enhance the practical utility of SFDI for real-time biomedical applications.

Main Methods:

• A deep learning-based inverse model was developed to process multi-frequency SFDI data.
• The deep learning approach was compared against conventional inversion algorithms.
• Performance was evaluated based on speed and accuracy of extracted optical property maps.

Main Results:

• The deep learning solution achieved speeds 300× to 100,000× faster than existing methods.
• The model demonstrated equivalent or superior accuracy in optical property estimation.
• The method effectively addresses the speed limitations of multi-frequency SFDI.

Conclusions:

• Deep learning provides a transformative solution for accelerating multi-frequency SFDI.
• This advancement facilitates real-time, highly accurate tissue optical property measurements.
• The proposed method is poised to significantly impact biomedical imaging research and clinical applications.