This dataset accompanies the research article entitled, "Etiology-Specific Remodeling in Ventricular Tissue of Heart Failure Patients and its Implications for Computational Modeling of Electrical Conduction," where we quantified fibrosis and performed electrophysiological simulation to investigate electrical propagation in etiologically varied heart failure tissue samples. Included are raw confocal microscopic images, data for extracting and processing the raw images and script to analyze fibrosis and generate meshes for simulation.
Significance: Current medical imaging systems have many limitations for applications in cardiovascular diseases. New technologies may overcome these limitations. Particularly interesting are technologies for diagnosis of cardiac diseases, e.g. fibrosis, myocarditis, and transplant rejection.
Aim: To introduce and assess a new optical system capable of assessing cardiac muscle tissue using light-scattering spectroscopy (LSS) in conjunction with machine learning.
Approach: We applied an ovine model to investigate if the new LSS system is capable of estimating densities of cell nuclei in cardiac tissue. We measured the nuclear density using fluorescent labeling, confocal microscopy, and image processing. Spectra acquired from the same cardiac tissues were analyzed with spectral clustering and convolutional neural networks to assess feasibility and reliability of density quantification.
Results: Spectral clustering revealed distinct groups of spectra correlated to ranges of nuclear density. Convolutional neural networks correctly classified 3 groups of spectra with low, medium, or high nuclear density with 95.00±11.77% (mean and standard deviation) accuracy. The analysis revealed sensitivity of the accuracy to wavelength range and subsampling of spectra.
Conclusions: LSS and machine learning are capable of assessing nuclear density in cardiac tissues. The approach could be useful for diagnosis of cardiac diseases associated with an increase of nuclei.
