Scar shape analysis and simulated electrical instabilities in a non-ischemic dilated cardiomyopathy patient cohort

Gabriel Balaban, Brian Halliday, Wenjia Bai, Bradley Porter, Carlotta Malvuccio, Pablo Lamata, Christopher A. Rinaldi, Gernot Plank, Daniel Rueckert, Sanjay K. Prasad, Martin J. Bishop

Non-ischemic dilated cardiomyopathy (NIDCM) is a significant clinical challenge characterized by the stretching and weakening of the heart's left ventricle. A primary concern for patients with NIDCM is the development of fibrotic scarring, which serves as a substrate for reentrant arrhythmias—the leading cause of sudden cardiac death. While Late Gadolinium Enhanced (LGE) MRI is the standard for identifying these scars, its resolution is often insufficient to capture the intricate "microstructure" of collagen fibers that dictates electrical behavior.

A study published in PLOS Computational Biology addresses this gap by combining clinical imaging with high-resolution electrophysiological simulations to identify macro-level "shape signatures" that predict lethal arrhythmias.

Methodology: Bridging Macro and Micro Scales

The researchers developed a robust computational pipeline to analyze 699 2D LGE-CMR image slices from a cohort of 157 patients. The study followed a four-step approach:

  1. Morphological Feature Extraction: Five key macro-level features were extracted from each MRI-detected scar: relative area, transmurality (the extent of the scar across the heart wall), interface length (the perimeter where scar meets healthy tissue), fragmentation (number of components), and entropy (textural disorder).
  2. Generation of Virtual Microstructures: To overcome MRI resolution limits, the team populated the scar regions with ten different hypothetical microstructures. These included interstitial and replacement fibrosis at five varying density levels, creating a library of high-resolution electrical environments.
  3. Electrophysiological Simulation: Using the CARP software package, 6,990 simulations were conducted. The team applied a programmed electrical stimulation protocol to observe whether specific scar shapes resulted in a transmural block (signal cessation) or transmural reentry (the signal looping back on itself).
  4. Statistical Modeling: Linear Discriminant Analysis (LDA) was employed to determine which macro-level shapes visible on standard MRI could best predict the electrical instabilities observed in the high-resolution simulations.

Key Findings: The Power of Interface Length

The study's most significant breakthrough is the identification of "interface length" as a critical predictor of arrhythmia. A longer, more complex border between the scar and healthy tissue increases the likelihood of "source-sink" mismatches. In these scenarios, a small electrical signal attempts to excite a larger area of tissue, leading to the slow conduction necessary to sustain a reentrant circuit.

Predicting Electrical Failure Modes

  • Transmural Block: Transmurality was the strongest individual predictor for electrical block (AUC = 0.79). The data suggests that if a scar does not extend across at least 20% of the heart wall, blocks and reentries are rarely observed.
  • Transmural Reentry: The most effective predictor for reentry was a combination of interface length, transmurality, and the number of components (AUC = 0.80). Notably, the LDA model showed that the coefficient for interface length was nearly six times greater than that for transmurality, suggesting that the "edginess" of a scar is far more dangerous than its depth alone.
  • Fragmentation Limits: Interestingly, the study found a limit to the danger of fragmentation. Reentry was not observed in scars broken into more than 15 separate components, suggesting that extreme fragmentation may actually "defuse" the risk of large-scale reentrant circuits.

Discussion and Clinical Implications

The research highlights that the macro-level shape of a scar, visible on a standard clinical scan, can act as a reliable proxy for the underlying microstructure that remains invisible to the naked eye. While the type of fibrosis matters—with high-density replacement fibrosis favoring blocks and lower-density interstitial fibrosis favoring reentry—the overall shape provides a powerful statistical indicator of risk.

Limitations and Future Directions

While the use of 2D simulations allowed for a high-volume analysis across a large patient cohort, the researchers acknowledge the limitations of a 2D model in representing a 3D organ. Future work aims to transition these findings into 3D "virtual heart" models, which could account for wavefront curvature and fiber rotation. Ultimately, these "shape signatures" could lead to personalized clinical tools, allowing doctors to better identify which NIDCM patients are at the highest risk for sudden cardiac death based on the specific geometry of their heart scars.

Summary updated

Abstract

Author summary Non-ischemic dilated cardiomyopathy is a disease in which the lower left chamber of the heart is abnormally large. The cause of the disease can be anything that is not a loss of blood supply to the heart. Many patients with non-ischemic dilated cardiomyopathy have scars in their hearts which can be detected with magnetic resonance imaging. These scars are thought to disrupt the flow of electricity through the heart and cause deadly rhythm disorders. In our study we look at various ways to quantify the shapes and textures of the scars with MRI images, and relate these metrics to rhythm disturbances that we simulate in image-based biophysical models of electrical activity. The exact electrical properties of scars in non-ischemic dilated cardiomyopathy are unknown, so our simulations account for a variety of possible properties. Our findings suggest new ways of examining scar shapes which work well with all of the properties we tested. This makes our shape metrics potentially valuable for determining which non-ischemic dilated cardiomyopathy patients are at risk of future heart rhythm disorders.
BibTeX
@article{balaban2019scar,
  author = {Gabriel Balaban and Brian Halliday and Wenjia Bai and Bradley Porter and Carlotta Malvuccio and Pablo Lamata and Christopher A. Rinaldi and Gernot Plank and Daniel Rueckert and Sanjay K. Prasad and Martin J. Bishop},
  title = {{Scar shape analysis and simulated electrical instabilities in a non-ischemic dilated cardiomyopathy patient cohort}},
  journal = {PLOS Computational Biology},
  year = {2019},
  doi = {10.1371/journal.pcbi.1007421},
}
Figure