Optical coherence tomography-based in vivo automatic detection of brain tissue malignancy using optical attenuation coefficient and refined speckle contrast: preliminary results

Optical coherence tomography (OCT) is a powerful non-invasive imaging technology; however, its conventional structural scans often suffer from limited contrast, hindering efficient differentiation of normal and pathological tissues. In this letter we increase the performance by extracting and combining two contrast-enhancing OCT signal parameters: optical attenuation and refined speckle contrast (SC) with attenuation correction. The depth-resolved estimation of the optical attenuation coefficient (OAC) helps to reveal the intrinsic scattering properties of the tissue. The SC estimation with mitigation of attenuation-induced inhomogeneities accentuates subtle microstructure-induced textural differences. We apply these two parameters to differentiate three classes of patients’ brain tissue using in vivo acquired OCT scans: (i) normal tissue (five patients, seven scans), (ii) non-malignant damaged tissue (four patients, five scans), (iii) malignant tissue (eight patients, 11 scans). We show that due to accentuated inhomogeneities in malignant tissues, SC becomes significantly informative for separation of normal and malignant tissues; however, for better diagnostic accuracy, it still needs to be combined with OAC. The individual SC and OAC performances have an accuracy of 61% and 78%, respectively, while in two-dimensional parametric space the delineation of the three considered classes (with linear support-vector machine-based clustering) provides an overall accuracy of 83%. However, it is important that 0% of true normal cases are misclassified as malignant and 0% of true malignant cases are misclassified as normal. Thus, for the available data volume, normal and malignant tissues were 100% correctly separated and the imperfect overall performance is only due to a few cases of misclassified malignant versus damaged normal tissue, as well as misclassified normal versus damaged normal. These promising results demonstrate the high potential of OAC + SC OCT for intraoperative separation of true malignant brain tissue from true normal tissue.