More than 20,000 hematopoietic stem cell transplants (including bone marrow transplants) are performed in the U.S. each year to cure a range of diseases ranging from leukemias to sickle cell anemia to autoimmune deficiencies in children. Unfortunately, most long-term non-relapse survivors will die of chronic graft-versus-host disease (cGVHD), which remains a disease of steadily increasing incidence and profound unmet need. A fundamental barrier in cGVHD management and research is a lack of sensitive and objective assessment tools that permit objective and reproducible measures of disease severity and progression. Skin is the most commonly affected organ in cGVHD and automated techniques capable of measuring precisely the surface area of involved skin in photographs may provide the tools necessary for effectively evaluating patient progress. We propose to (1) create the data set necessary to develop machine learning-based methods for the automatic analysis of cGVHD images, and (2) implement and evaluate these methods.
Chronic graft-versus-host disease (cGVHD) is a lethal disease that affects most long-term hematopoietic stem cell transplant (including bone marrow transplants) recipients. Skin images are used to assess disease severity and progression but the technology required to quantitatively and reproducibly analyze these images is lacking. This project aims at developing and evaluating this technology.
