Conventional testing for SLD has relied on performing biopsies, in which a tissue sample is taken from the liver for analysis. The method developed by TUS professor Kohei Soga and his team uses NIR light to visualize lipid content in the liver. NIR wavelengths can be used to identify fat distribution in the liver because they are long enough (800 to 2500 nm) to reveal the absorption of biomolecules in deep tissues.
However, the team found that, while NIR-HSI could map the distribution of whole lipids, it could not provide the ability to visualize various properties in lipids, such as molecular weight and single or double bonds.
The imaging framework visualizes hydrocarbon chain length and degree of saturation of fatty acids in mice livers by combining near-infrared hyperspectral imaging and machine learning. Courtesy of Mori et al. (2023). Scientific Reports, doi: https://doi.org/10.1038/s41598-023-47565-z.
To resolve this issue, the TUS team, working with researchers at Osaka Metropolitan University, adopted a support vector regression (SVR) machine learning model and trained the model to recognize the composition of 16 fatty acids found in the liver. The researchers acquired the training data through gas chromatography analysis of liver samples of mice.
By applying machine learning to the NIR-HSI data, the researchers were able to interpret the spectral information and use it to analyze the distribution of the hydrocarbon chain length (HCL) and degree of saturation (DS) of fatty acids within the mice livers. The total lipid content in the tissues, as well as the structural characteristics of the fatty acids, could be visualized from the NIR reflectance spectra of the tissues. The machine learning model differentiated the type of lipids present in the liver at a pixel-by-pixel level.
“In addition to quantitative information, such as the total lipid content, we can now also visualize qualitative information, such as the characteristics of the distribution of fatty acids contained in lipids, mainly triglycerides,” said TUS professor Masakazu Umezawa.
The researchers performed a 2D mapping of the HCL and DS of fatty acids in the mice livers to determine the fatty acid composition. They categorized the 16 fatty acids based on HCL and DS. By analyzing NIR (1000–1400 nm) spectra acquired using SVR, the researchers were able to predict the average values of the HCL and DS of fatty acids for each lobe of the mouse liver, in addition to the total lipid concentration.
The team found a correlation between the fatty acid distribution and the fat contents in the diets of the mice. The livers of mice on a diet rich in saturated fats exhibited a high DS, whereas mice fed with unsaturated fats had a low DS.
The DS, HCL, and total lipid content of the mice livers were depicted as a color map, providing a comprehensive visual representation of fat distribution in the livers. In the future, a visual like this could simplify the diagnosis of fatty liver conditions.
“Visualization of lipid distribution in higher-dimensional information rather than simply using total lipid content as a single parameter provides a novel tool for revealing the pathophysiological conditions of liver diseases and metabolism,” Umezawa said.
Looking ahead, NIR-HSI could be incorporated into a laparoscope as an alternative to liver biopsy to assess a patient’s risk of SLD progression, steatohepatitis (NASH), and SLD/NASH-associated liver cancer. The technique’s ability to non-invasively map the distribution of fatty acids with their chemical structures removes the need for slicing, homogenizing, or chemically staining organ samples.
The NIR-HSI framework could also be used in pharmacological research — for example, to study drug metabolism, toxicity, and efficacy and in studies on metabolic disorders through metabolic imaging. Additionally, it could help identify personalized nutrition strategies, tailor nutrition plans, and optimize interventions for better nutrition, by identifying biomarkers and predicting treatment response.
The research was published in Scientific Reports (www.doi.org/10.1038/s41598-023-47565-z).
