Using machine learning, biomedical engineers have developed a technique to improve optical coherence tomography (OCT) resolution to a single micrometer in all dimensions, even in a living patient. The multibillion-dollar optical coherence tomography (OCT) market could benefit from the new optical coherence refraction tomography (OCRT) method by getting better medical images for fields like oncology and cardiology.
Researchers use a distinct strategy in the most recent paper. The researchers blend OCT images obtained from various angles instead of holography to increase the depth resolution to the lateral dimension. However, due to the light’s refraction through imperfections in the cells and other tissue components, each unique OCT image becomes distorted. The researchers had to precisely model how the light is bent as it passes through the sample to account for these modified paths when assembling the finished images.
Researchers created a “gradient-based optimization” technique using multi-angle images to determine the refractive index within the various tissue areas. This method decides which direction the given property—in this instance, the refractive index—needs to be adjusted to produce a better image. A map of the tissue’s refractive index is produced by the machine learning algorithm after numerous trials to best account for the distortions of the light. TensorFlow, a well-known software library developed by Google for deep learning applications, was used to put the technique into practice.
Related Content: PR-OCT Machine Learning For Colon Cancer Detection