Portrayal of Cancer Patients in the Era of AI

A Content Analysis of Images Produced by Generative AI Tools

Bibliographic Data

ID21617055
AuthorsWen-Ying Sylvia Chou (0000-0002-9140-6094, National Cancer Institute, corresponding author), Anna Gaysynsky (0000-0001-5612-5572, National Cancer Institute), Nicole Senft (0000-0002-3617-7276, National Cancer Institute), Abigail Muro (0009-0008-9080-2416, National Cancer Institute), Kristin Schrader (Westat), Irina A Iles (0000-0002-9348-1012, National Cancer Institute)
Year2026
Volume41
Issue5
Pages888-898
Publication date2026-04-16
Peer ReviewedYes
Open AccessNo
TypeARTICLE
VenueHealth Communication (JOURNAL)
Journal identifiersISSN: 1041-0236 • E-ISSN: 1532-7027
PublisherInforma UK Limited (PUBLISHER • GB)
DOI10.1080/10410236.2025.2537807
PMID40776397
OpenAlexW4413051479
LanguageEN
References cited47

This study sought to characterize images of cancer patients generated by Artificial Intelligence (AI) text-to-image tools, and assess whether images differed by cancer type or AI tool, to elucidate the potential implications of using AI-generated images in health communication. Two generative AI-based tools, DALL-E and Stable Diffusion, were prompted to produce images of a “cancer patient,” “breast cancer patient,” “lung cancer patient,” and “prostate cancer patient”. Images (N = 320) were coded for perceived demographics, illness features, affect, cancer symbols, setting, and photorealism. Analysis revealed that AI tools commonly depicted cancer patients as White (83.2%) and middle-aged or older (87.5%). Compared to general cancer patient images, breast cancer patients were portrayed as younger, while prostate and lung cancer patients were depicted as older. Breast cancer patients were also more frequently depicted as healthy and displaying positive affect, while lung cancer patients were more often depicted as ill and showing negative affect. Differences were also found between the AI tools, with DALL-E images featuring more racial diversity and being less photorealistic compared to images produced by Stable Diffusion. Because generative AI tools may produce images of cancer patients that are limited on some dimensions of diversity, and in some cases may reinforce stereotypes (eg, breast cancer patients as healthy and happy, lung cancer patients as ill and hopeless), it is critical to consider biases that may exist in these models — and the potential societal implications of using AI-generated images of cancer patients — before these tools are deployed in cancer communication efforts

Breast cancer · Cancer · Lung cancer · Prostate cancer · Communication · Empathy and Medical Education · Media Influence and Health · Medicine · Patient-Provider Communication in Healthcare · Psychology · Internal Medicine · Oncology

Tools

Open DOI