Optical Insights into Fibrotic Livers: Applications of Near-Infrared Spectroscopy and Machine Learning

Application of NIR and machine learning for fibrotic liver analysis © PIC4U- stock.adobe.com

Liver fibrosis, a common outcome of chronic liver diseases such as hepatitis and alcohol abuse, is a major global health concern. If left untreated, it can progress to cirrhosis or liver failure. The worldwide occurrence of cirrhosis is not known but is estimated to be approximately 0.21% of the population in the United States (1). Assessing the degree of fibrosis is crucial, especially for transplant patients, yet traditional methods like biopsy are invasive and often inaccurate. In this inventive study, researchers have proposed the use of near-infrared spectroscopy (NIRS) coupled with machine learning (ML) to create a more efficient, non-invasive solution for liver fibrosis staging (2).

The Need for Better Liver Fibrosis Detection

Liver fibrosis progresses as the body accumulates extracellular matrix proteins, including collagen, in response to liver injury. These changes significantly affect liver tissue composition, leading to the need for better methods of staging the disease. Currently, liver biopsy remains the gold standard, but it has limitations, including risk of complications and sampling errors. Non-invasive biomarkers and imaging techniques, such as the aspartate aminotransferase (AST) to platelet ratio index (APRI) and transient elastography, offer alternatives but lack accuracy, especially for intermediate stages of fibrosis (2).

NIRS and Its Role in Liver Fibrosis Detection

NIRS offers a promising alternative. This optical technique detects changes in light absorbance patterns, which vary depending on the chemical composition of tissues. In fibrotic livers, increased collagen content and alterations in lipid, protein, and water content create distinct spectroscopic signatures. Researchers have shown that NIRS can reliably distinguish between fibrotic and healthy liver tissues by analyzing these patterns (2).

This study by Tamer A. Addissouky, Ibrahim El Tantawy El Sayed, and others from Al-Hadi University College in Baghdad, Iraq; Menoufia University in Menoufia, Egypt; and MLS Ministry of Health in Alexandria, Egypt has been published in the journal Archives of Gastroenterology Research. The work involves tissue samples from explanted and unused donor livers, and found that NIRS could predict liver fibrosis stages with impressive accuracy. The research demonstrated that the technique could differentiate mild from advanced fibrosis, as well as detect cirrhosis, with high sensitivity and specificity. NIRS was able to capture molecular and structural changes associated with liver fibrosis, such as increased collagen and changes in lipid content (2).

Integrating ML with NIRS for Diagnostic Models

While NIRS provides rich data, interpreting this information requires advanced computational techniques. ML algorithms like partial least squares regression PLS, support vector machines (SVM), and neural networks (NNs) have been applied to NIRS spectra to build predictive models. These ML techniques can analyze complex data sets, identify patterns, and generate accurate diagnostic algorithms for staging liver fibrosis (2).

In this study, a combination of ML algorithms was used to analyze NIRS data, correlating it with histological findings from liver tissue samples. The results showed that models trained on NIRS spectra could accurately classify fibrosis stages, even outperforming traditional biomarkers like aspartate aminotransferase-to-platelet ratio index (APRI) and fibrosis-4 (FIB-4). These ML models offered superior diagnostic accuracy for differentiating between early and advanced stages of fibrosis (2).

Technological Innovations in NIRS-ML Systems

Recent advancements in technology have further enhanced the potential of NIRS for liver fibrosis assessment. Handheld and hyperspectral imaging (HSI) systems now allow for quick tissue analysis in clinical settings. Smartphone-based scanning and cloud computing have made these systems more accessible, with platforms that offer real-time data analysis. However, challenges remain in optimizing these systems for widespread clinical use. Factors like probe pressure, skin color, and body mass index (BMI) may impact NIRS readings, and further research is needed to account for these variables (2).

Promising Results and Current Challenges

The combination of NIRS and ML shows great promise, as seen in various studies. One study demonstrated that a random forest classifier using NIRS data could differentiate severe fibrosis from mild disease with an accuracy of 82%, outperforming traditional biomarkers. Another study showed that NIRS, when combined with neural networks, achieved 100% sensitivity and 89% specificity in detecting advanced fibrosis. Despite these encouraging results, further validation is required to ensure accuracy across diverse patient populations (2).

Challenges remain in standardizing the methodology and addressing confounding factors, but ongoing research is focused on improving ML algorithms, refining data collection techniques, and making NIRS-ML systems more user-friendly and cost-effective (2).

Future Directions and Recommendations

To translate NIRS-ML technologies into routine clinical practice, further large-scale studies are necessary. Multi-center trials will help assess the accuracy and reliability of these systems in diverse populations. Additionally, optimizing ML techniques for improved generalizability and interpretability is crucial (2).

Researchers also recommend that the industry focus on developing portable, point-of-care devices that can provide real-time results in various clinical environments. Overcoming the remaining barriers, such as standardizing protocols and accounting for patient variability, will help make NIRS-ML systems a routine tool in the assessment of liver fibrosis, particularly in transplant surgery planning (2).

The combination of near-infrared spectroscopy and ML holds significant promise for the rapid, non-invasive detection of liver fibrosis. By improving the accuracy, convenience, and accessibility of fibrosis staging, this approach could transform the management of liver disease and revolutionize the field of transplant surgery. With ongoing advancements and further validation, NIRS-ML technology could become the new gold standard for assessing liver fibrosis in clinical settings. As technology advances, the marriage of NIRS and machine learning is paving the way for a future where liver fibrosis can be diagnosed with precision and ease, bring improvements to the clinical landscape.

References

(1) NIH Hepatic Cirrhosis Web Page.https://www.ncbi.nlm.nih.gov/books/NBK482419/ (accessed 2025-01-20).

(2) Addissouky, T. A.; El Sayed, I. E. T.; Ali, M. M.; Alubiady, M. H. S. Optical Insights into Fibrotic Livers: Applications of Near-Infrared Spectroscopy and Machine Learning. Arch. Gastroenterol. Res. 2024, 5 (1), 1–10. 10.33696/Gastroenterology.5.048.