Transflection Interference Influence On Spectral Quality And Histopathologic Classification Of Infrared Imaging Data

Infrared (IR) Imaging can be used for fast, accurate and non-destructive pathology recognition of biopsies when supported by machine learning algorithms. Transflection mode of measurements is the most probable one to be translated into the clinic due to economic reasons of large scale imaging with expensive transmissive substrates (CaF2, BaF2). A typical procedure after collection of a biopsy from the patient is embedding the sample within paraffin wax. This has an impact on sample optical properties on one hand minimizes scattering, but on the other increases interference (so called Electric Field Standing Wave effect) and distorting spectra, by creating a flat, thin film. Moreover, interference has been shown to be more prominent in a coherent IR source, such as Quantum Cascade Laser (QCL) [2] and is expected to be also high using a synchrotron source. The question, whether lower scattering but higher interference in a paraffin-embedded sample will provide better spectral quality over dewaxed sample, with higher scattering and lower interference, remains open. In this work we will investigate this aspect by evaluating classification performance of a Random Forest classification of two tissue types (pancreas, esophagus) measured in transmission/transflection, FT-IR vs two QCL microscopes and with and without paraffin.