Invasive biopsies are critical in the medical diagnosis and characterization of diseased tissues, but are hindered by risk of infection, tissue damage, and high post-processing time. Optical approaches for tissue characterization allows for faster and less invasive procedures [1]. One promising optical approach is light-scattering spectroscopy (LSS) [2, 3]. Using LSS, we distinguished tissue based on properties like nuclear density [4], and tissue composition [3]. However, questions remain about the efficacy of LSS in vivo, and its clinical applications. A potential use of LSS is identifying the cardiac conduction system, e.g. to avoid damaging during surgical repair of congenital defects. Here, we describe a study on the in situ, beating heart in canines to provide insight into the capabilities of LSS.
Spectra were gathered in vivo from four adult canine models using a catheterized LSS probe [3,4]. 10 samples of 200 spectra were gathered from the atrium, ventricle, vena cava, and blood. Data analysis was performed in Matlab r2021a. The spectra were calibrated, normalized, and averaged across each set of 200 spectra. principal component analysis (PCA) and Uniform Manifold Approximation and Projection (UMAP) were used to reduce dimensionality [5]. ANOVA, with a Tukey-Kramer post hoc test and a significance level of 0.05, was used to identify differences between the first principal components of the different tissue regions [6]. The same approach was used with the first UMAP index.
The results of the PCA and UMAP are shown in figure 1. The first two principal components explain 66.3% and 21.9% of the variance respectively. ANOVA using the first principal component yielded differences (p<0.05) between all groups except between the atrium and ventricle (p=0.86). The UMAP shows more visible separation between tissues, and ANOVA yielded every group to be statistically different using the first index (p<0.05). These results demonstrate the ability of the spectra to differentiate between tissue. The relative similarity of spectra from the atrium and ventricle was expected because both regions consist of muscle cells, which vary structurally from the blood and vena cava.
We demonstrate the feasibility of using LSS in vivo to non-destructively characterize tissue regions in the heart. UMAP identified tissue-specific clustering in the first index, indicating that key information about tissue properties lie in the spectra. This study informed future studies for our LSS system. Further work using supervised learning approaches and heterogenous tissue samples will provide more insight into the capabilities of LSS for tissue discrimination [4,5]. LSS' ability to differentiate tissue based on composition, both ex and in vivo, facilitates translation for medical applications. Our studies suggest that LSS has promise as a means of diagnosing and characterizing tissues in the heart, thereby improving our ability to quickly identify and effectively treat diseases.