Deborah A Striegel, PhD

Sepsis heterogeneity makes it difficult to predict which patients will survive the acute phase of critical illness and of those who do survive, which patients will develop long term sequalae. Our work at ACESO has focused on defining endotypic states of immune dysfunction, such as immunosuppressed, immunometabolic, acute inflammation, and immunocompetent states (Chenoweth, et al., 2024) and changes in immune profiles over time emphasizing paths returning to a healthy state (Chenoweth, et al., 2024).

Determining a patient's path after critical disease insult is essential for personalized treatment. At ACESO we have developed a trajectory clustering algorithm, Dynamical Systems Cost-fitting Landscapes, that utilizes dynamical systems modeling with machine learning to cluster patient trajectories.

Standard spectral clustering relies on a single global scale parameter, failing to capture complex datasets with nested, multiscale, or varying-density cluster structures. At ACESO, we have developed an algorithm, Multi-scalar Spectral Clustering, that utilizes the global scale parameter to find scale-dependent clusterings within a dataset and the persistance of the clusterings.