Random matrix theory to quantify micro-structural changes in rodent lungs due to pulmonary diseases

We exploit the random matrix theory to detect changes in rodent lungs exhibiting pulmonary fibrosis and edema. Coherences in the backscattered signals are stronger when single scattering dominates (fibrosis/edema). On the contrary, healthy lungs exhibit more apparent randomness due to multiple scattering. This leads to differences in the distribution of eigenvalues, which can be retrieved using Singular Value Decomposition of the Inter-elementResponse Matrix (IRM). We use features of the eigenvalue distribution (E(x), the expected value, and, the eigenvalue with the highest probability) to quantify changes in lung parenchyma and investigate whether they can improve the specificity of quantitative ultrasound to lung diseases. IRMs were acquired from 51 rat lungs (10 controls, 18 edematous, 17 fibrotic, 6 fibrotic rats, which were treated with Nintedanib) using a 128-element linear array (Verasonics L11-4v, 7.8 MHz). Severity of fibrosis and edema were quantified by histology and the ratio of wet to dry weight. Both parameters showed significant differences between edematous and fibrotic lungs, and between control and fibrotic lungs, which was significantly correlated to both the severity of fibrosis and edema. E(x) was significantly correlated to the severity of fibrosis. This suggests that these parameters could be part of a toolkit for the quantitative assessment of lung diseases.