Methods for the Estimation of Depth-resolved Attenuation in OCT

Cancer is one of the major causes of mortality, accounting for one in six deaths worldwide in 2020. The majority of cancers can be prevented , cured , or reversed if early detected. Early diagnosis and screening are vital components, with screening focusing on detecting cancer before the symptoms manifest, and diagnosis involving the identification of cancer in its early stages through methods like biopsies and imaging techniques.
Optical coherence tomography (OCT) has shown potential for early cancer diagnosis, due to its non-invasive nature and potential to guide biopsies site selection, reducing the invasiveness of certain procedures by enabling the clinicians to obtain fewer biopsies and minimizing errors associated with surgical biopsies. Researchers have proposed several models linking the attenuation coefficient (μ), which describes the exponential decay of the OCT signal, with tissue optical structural properties. Presently , two methods, layer-wise extraction through curve fitting and depth-resolved methods, are employed for attenuation coefficient calculation. However, challenges persist in achieving accurate and precise estimations.
This thesis entails a comparison of four methods to calculate the attenuation coefficient from OCT intensity data-: slope-fitting, Vermeer, Jain Lui and Kaiyan Li. Validation of theses methods on various types of milk serves as a preliminary evaluation, with further comparisons of estimated attenuation coefficients from OCT images against measurements from a power meter using various intralipid concentrations. The identified optimal algorithm is subsequently applied to OCT images for both cancerous and normal oral tissues.
The study extends its evaluation to explore critical factors influencing the performance of the four methods. Pre-processing, including frames averaging, is found to enhance data quality and signal-to-noise ratio (SNR). The noise floor, particularly prominent at greater imaging depth, poses challenges common to all methods. Overlooked confocal effects, especially in samples with lower attenuation coefficients, contribute to underestimation in the Jain Lui and Kaiyan Li methods. While the validation experiments establish a linear relationship between intralipid concentration and attenuation coefficient, multiple scattering at higher intralipid concentrations challenges the assumption of constant backscattering fraction. Each method exhibits strengths and shortcomings, emphasizing the need for further research to explicitly incorporate confocal effects and multiple scattering to enhance depth-resolved algorithms for improved biomedical applications.