Chronic graft-versus-host disease (cGVHD) is the leading cause of nonrelapse mortality following allogeneic hematopoietic stem cell transplantation (HCT). Once the diagnosis is made, a fundamental practice gap remains the determination of whether disease is stable or progressing. Clinical trials of promising new potential treatments are limited by the lack of reproducible and sensitive measures of cGVHD severity. Skin is central to cGVHD evaluation because it is the most commonly involved organ. Features are divided into erythema (visualized changes) and sclerosis (palpated mechanical changes). This project will implement an objective, longitudinal monitoring approach to cGVHD by combining 3D digital photography, machine learning, and biomechanical assessment with the Myoton device. These technologies look at and feel skin analogously to a clinical exam, but in a precise and quantitative fashion. The hypothesis of the proposed research is that this integrated technological approach will reliably detect clinically important changes in disease severity. This will provide the opportunity to overcome the shortcomings in existing methods, enabling quantitative assessments to validate and guide therapy.
Aim 1 will test the reliability and reproducibility to quantify erythema body surface area with 3D photography and deep learning. A large patient image data set will be created to optimize and test the reliability of a deep learning convolutional neural network to independently identify, demarcate and grade regions of erythema.
Aim 2 will test the reproducibility of biomechanical assessment of skin sclerosis with the Myoton, a handheld commercial device that is widely used to noninvasively measure biomechanical and viscoelastic properties of muscle.
Aim 3 will evaluate the ability of the integrated quantitative approach to measure clinically meaningful changes in cGVHD severity in a year of follow- up of a prospective cohort of cGVHD patients. The proposed neural-network assessment of erythema and skin biomechanical assessment with Myoton are each significant innovations, which can later be applied to a broad range of other progressive cutaneous diseases. The proposed work is significant because it addresses the inability to accurately measure cGVHD severity and treatment response, which is currently the fundamental barrier to permanent successful treatment by HCT of hematologic malignancies and other diseases.
Veterans are at increased risk for hematopoietic system diseases like leukemia. The only cure for many patients with these diseases is hematopoietic stem cell transplantation (HCT). However, the new immune system from the transplant (graft) often launches a multisystem attack against the patient (host), termed chronic graft-versus-host-disease (cGVHD). Treatment options are inadequate, and cGVHD will kill most long- term HCT survivors who overcome their original cancer. Clinical trials of promising new treatments as well as the ability to make optimal patient care decisions are thwarted by the absence of assessments to determine whether patients are responding to treatment or progressing. The proposed research will study the ability to objectively and reliably measure cGVHD in skin by a combination of innovative handheld, noninvasive tools: 3D photography, artificial intelligence, and the Myoton device. This may enable evidence-based, personalized cGVHD therapies and guide treatment decisions for improved outcomes in Veteran and other HCT survivors.