Early cancer diagnosis is critical for improving treatment outcomes and lowering mortality rates. However, prompt diagnosis is common for people who need access to healthcare, such as those with limited income or living in remote areas. Diffuse reflectance spectroscopy (DRS) is a promising method for early cancer diagnosis since it may be conducted rapidly and with low-cost equipment.
Inverse Monte Carlo (MCI) simulations are widely regarded as the gold standard for evaluating DRS data and determining tissue optical characteristics. However, when applied to real-world measurements, ML-based algorithms could perform better. A research team created a more robust machine learning model dubbed “wavelength-independent regressor” (WIR) to analyze DRS data and forecast the absorption coefficient (μa) and the reduced extinction coefficient (μ′s). The proposed model employs a novel collection of DRS data features to attain higher accuracies in the face of use mistakes.
The researchers employed a large dataset that included simulated data and experimental observations collected from 170 tissue phantoms. The WIR model achieved the optimum balance of accuracy and speed. In experimental data, the WIR model showed mean errors of 13.2% and 6.1% for μa and μ′s, respectively, while the errors for MCI were approximately eight times larger.
The method developed in this study will help with early cancer diagnosis and other diseases. The suggested strategy is less computationally intensive than previous ML-based techniques and MCI models, making it a viable option in clinical settings.
Related Content: Infrared Sensor Uses AI, Plasmonics To Detect Neural Diseases