Precision health for all: developing inclusive datasets and algorithms

BIODS 260
Roxana Daneshjou
10/06/22
1:30 pm-2:30 pm
Dr. Roxana Daneshjou

Location: MSOB x303 

Seminar Title: Precision health for all: developing inclusive datasets and algorithms

Abstract: Large biomedical datasets coupled with machine learning tools have the potential to transform the practice of dermatology. For example, analysis of skin disease images could help triage patients prior to the clinical visit and precision genomic medicine could identify personalized treatments for skin disease. However, biased datasets and algorithms that exclude underrepresented groups could exacerbate existing health disparities in dermatology. This talk will discuss working towards inclusive precision medicine through three examples: 1) assessing fairness in datasets and AI algorithms used for diagnosing disease in dermatology 2) developing an inclusive patient-facing algorithm to improve the quality of images submitted for teledermatology and 3) developing a pharmacogenomics algorithm that accounts for population diversity.  In order to develop a data-driven approach to dermatology that improves health disparities, rather than exacerbating them, we must be mindful of developing inclusive datasets and algorithms.

Suggested Readings:

• https://www.science.org/doi/10.1126/sciadv.abq6147
• https://jamanetwork.com/journals/jamadermatology/article-abstract/2784295
• https://ashpublications.org/blood/article/124/14/2298/33101/Genetic-variant-in-folate-homeostasis-is