Dr. Valeria Grasso describes a novel unsupervised Machine-Learning method named Superpixel Photoacoustic Unmixing (SPAX) framework.
The SPAX enables the data-driven differentiation of molecular tissue components (such as oxy/deoxy hemoglobin, melanin, lipid, collagen, etc.) from spectral Photoacoustic Imaging (PAI), without any user interactions or prior information.
The framework is also extended to compensate for the spectral coloring artifact that affects Photoacoustic Imaging accuracy at depth and furthermore, an advanced Superpixel subsampling process that increases the detection of less prominent tissue chromophores, those that are generally obscured from the highly absorbing molecules, has also been developed.
The SPAX also includes an optimized visualization of the molecular components’ distribution at a volumetric scale with unprecedented detail. Extensive analysis has been conducted to benchmark the performance of the formulated SPAX approach for endogenous and exogenous chromophores blind tissue characterization in vivo.
Finally, the SPAX is an advanced data-driven approach that has the potential to enhance the sensitivity and specificity of photoacoustic molecular image unmixing and thus open many possibilities to expand the applicability of PAI for preclinical and translational science.
Key Topics Include:
- Advantages of Unsupervised Machine Learning unmixing algorithms for spectral photoacoustic imaging
- SPAX key features for accurate and complete tissue characterization
- In vivo application of SPAX for non-invasive and automated tissue monitoring
Psychiatry & Behavioral Sciences and Radiology & Biomedical Imaging
University of California San Francisco