Machine Learning Models to Differentiate the Etiology of Cutaneous Reactions in Post-Stem Cell Transplant Patients
- Nimesh Nagururu
- Clara Lemaitre
- Vince Wang
- Audrey Lacy
- Nandita Balaji
- Jonathan Hung
- Joseph L Greenstein
- Casey Overby Taylor
- Sima Rozati
- David Weiner
- Austin Burns
- Olivia Pierog
Abstract:
Allogeneic hematopoietic stem cell transplantation (HSCT) is a life-saving treatment for patients with hematologic diseases and is often complicated by post-transplant cutaneous eruptions. The etiologies of these rashes include drug reactions, viral exanthema, and graft-versus-host disease (GVHD), which require timely and unique treatment regimens to lessen associated morbidity and mortality. Differentiation of post-HSCT rashes is challenging within current diagnostic paradigms. Therefore, there is a need for robust computational models that can integrate clinical and lab data to appropriately identify post-HSCT rash etiology. In this investigation, we employ supervised and unsupervised machine learning models ad a single-institution dataset to help distinguish GVHD and non-GVHD rashes.