AI and Myeloma: HealthTree Research Finds a Fixable Problem With M-Protein Data | IMS 2026

Ehsan Malek, MD, of Roswell Park Comprehensive Cancer Center discusses his research collaboration with HealthTree Foundation on using artificial intelligence in myeloma at IMS 2026 in Glasgow. Using patient-donated HealthTree records, the project found that about half of patients' monoclonal protein results are stored as text in medical records rather than as standard lab values, making them hard for AI to use. That gap could bias future AI tools toward higher-income patients. He explains why he sees this as a very fixable problem.

If this video has helped you or a loved one, consider donating to HealthTree. Every donation helps us continue to create free and high-quality blood cancer educational content: https://give.healthtree.org/give/628617/src=htu

For an organized lesson structure, watch on our site here: https://linktr.ee/HealthTreeUniversity

Create a free account to track your progress, bookmark lessons, and keep up-to-date on the latest advancements in treating cancer!

Patients have the key to find cures for cancer. Learn more about HealthTree's services here: https://healthtree.org/curehub?utm_source=social+media&utm_medium=htu+youtube&utm_campaign=htu-curehub&utm_content=general

Educated and empowered patients have better outcomes. We've partnered with hundreds of medical experts and doctors to help you learn more about your disease.

Timestamps:

0:00 Introduction: AI and myeloma

0:19 Why myeloma is hard for AI

0:56 What HealthTree data revealed: M-protein stored as text

1:57 Why fixing it matters

2:13 What "embedded in text" means

2:50 Socioeconomic bias in AI training data