Ophthalmology Interest Group 2/9 Lunch Lecture
Friday, February 9 from 12-1pm in MMEC 7-102.

Title: AMD Cell Therapy Efficacy Assessment Using Artificial Intelligence-Based Multi-Spectral Imaging

Speaker: Nathan A. Hotaling, PhD
Unit on Ocular and Stem Cell Translational Research, National Eye Institute, National Institutes of Health

Email *
Your answer
Name *
Your answer
Year *
Do you have any dietary requests? *
Do you have any questions for Dr. Hotaling?
Your answer
Age-Related Macular Degeneration (AMD) is a progressive blinding disease with no cure at present. Approximately 11 million people have AMD in the United States, and there is an estimated global prevalence of over 170 million. AMD is the leading cause of visual disability in the industrialized world and the third leading cause globally. Pharmaceutical approaches have been ineffective in treating the most prevalent form of AMD (dry AMD) and thus, cell therapy approaches are currently being developed. Dry AMD is believed to be mediated by the disfunction and death of Retinal Pigment Epithelial (RPE) cells. Therefore, cell therapies have targeted this cell type to treat the disease. However, to develop a commercially viable cell therapy product, consistent biomanufacturing of cells/tissues is necessary. The Food and Drug Administration (FDA) requires release criteria for cells/tissues that assess product potency, identity, and batch-to-batch variability. The assays currently used are invasive, labor intensive, time consuming, and/or highly variable. A new method that is non-invasive, automated, fast, and robust is needed for these therapies to clear regulatory hurdles and be commercially successful.

I will first present work showing that the biochemical and physiological function of induced pluripotent stem cell derived RPE (iPSC-RPE) can be predicted using novel multi-spectral imaging, machine learning algorithms, and convolutional neural networks (CNN). Then, I will show that this methodology was able to be used on iPSC-RPE derived from patients with dry AMD that were differentiated using an FDA compliant good laboratory practices (GLP) protocol. When combined, the imaging and computational algorithms were able to predict cell function with high sensitivity and specificity as well as classify iPSC-RPE identity to an equal degree as that of traditional physiological and molecular assays. Additionally, traditional machine learning approaches were able to identify critical visual components to predicting cell function. Thus, not only can visual information be correlated to cell function/identity, but also possible mechanisms of action can be identified using this approach. The process is fully automated, relatively inexpensive, and needs no human intervention to perform.

In summary, I show that the combination of multispectral imaging, machine learning, and CNNs are a powerful tool. This tool can not only be used to assess RPE cell function and identity in culture but also as a non-invasive release criterion for an iPSC-RPE cell based therapy to evaluate product potency, identity, and consistency.

Never submit passwords through Google Forms.
This form was created inside of Virginia Commonwealth University. Report Abuse - Terms of Service