Saak Transform-Based Machine Learning for Light-Sheet Imaging of Cardiac Trabeculation

Abstract

<jats:title>Abstract</jats:title><jats:p>Recent advances in light-sheet fluorescence microscopy (LSFM) enable 3-dimensional (3-D) imaging of cardiac architecture and mechanics <jats:italic>in toto</jats:italic>. However, segmentation of the cardiac trabecular network to quantify cardiac injury remains a challenge. We hereby employed “subspace approximation with augmented kernels (Saak) transform” for accurate and efficient quantification of the light-sheet image stacks following chemotherapy-treatment. We established a machine learning framework with augmented kernels based on the Karhunen-Loeve Transform (KLT) to preserve linearity and reversibility of rectification. The Saak transform-based machine learning enhances computational efficiency and obviates iterative optimization of cost function needed for neural networks, minimizing the number of training data sets to three 2-D slices for segmentation in our scenario. The integration of forward and inverse Saak transforms serves as a light-weight module to filter adversarial perturbations and reconstruct estimated images, salvaging robustness of existing classification methods. The accuracy and robustness of the Saak transform are evident following the tests of dice similarity coefficients and various adversary perturbation algorithms, respectively. The addition of edge detection further allows for quantifying the surface area to volume ratio (SVR) of the myocardium in response to chemotherapy-induced cardiac remodeling. The combination of Saak transform, random forest, and edge detection augments segmentation efficiency by 20-fold as compared to manual processing; thus, establishing a robust framework for post light-sheet imaging processing, creating a data-driven machine learning for 3-D quantification of cardiac ultra-structure.</jats:p>

A Hybrid of Light-Field and Light-Sheet Imaging to Decouple Myocardial Biomechanics from Intracardiac Flow Dynamics

“…We generated the myocardial displacement vector maps by exploiting the gray scale raw images, obviating the need for manual segmentation and annotation. While this enables frame-toframe analyses of the mechanics with high resolution, endocardiac segmentation would further enhance the accuracy of estimating the volumetric geometry needed to compute metrics such as ventricular ejection fraction (Akerberg et al, 2019;Ding et al, 2020).…”

“…We have previously established the multi-scale light-sheet imaging with superresolution to elucidate the initiation of trabeculation from zebrafish to mouse hearts (Ding et al, 2018;Fei et al, 2019). Using the time-dependent vector analysis and deep-learning approach (Chen et al, 2019), we further enhanced auto-segmentation and quantification of trabecular volume in relation to the compact zone (Ding et al, 2020;Lee et al, 2016;Vedula et al, 2017).…”