UCSD Researchers Innovate Skin Cancer Detection Using AI
Skin cancer is highly treatable when identified early, yet detection can vary depending on skin tone. Those with darker skin may receive delayed diagnoses because skin cancer presents differently across ethnicities. Traditional models often overlook these variances, leading to disparities in early detection.
The Issue of Late Diagnosis in Diverse Populations
Skin cancer manifests distinctively based on an individual's skin tone. For people of European ancestry, traditional detection methods are more reliable, but these methods are less effective for non-European ethnicities. As genome researcher Kelly Frazer from UC San Diego points out, cancer may appear in less visible or unexpected areas like under the fingernails or on the soles of the feet for people with darker skin tones. These non-typical presentations can lead to later detection and treatment.
The Role of AI in Addressing Disparities
Frazer and her team at UC San Diego have sought to address these disparities with the help of artificial intelligence (AI). Their research aimed at developing a tool that could predict skin cancer risk more accurately across diverse ethnic backgrounds. They leveraged data from over 400,000 individuals using whole genome sequencing, which provides comprehensive insights into a person’s DNA. This data was sourced from the National Institutes of Health's All of Us research program, which emphasizes diversity and inclusivity in its genetic database.
By integrating genetic data with various non-genetic factors such as income level, lifestyle choices, and medication use, the researchers developed an AI model capable of identifying individuals at high risk for skin cancer. According to Frazer, this model could become a vital tool for early identification, prompting full-body scans for those flagged by the AI.
The Impact of AI on Accuracy and Equality
In preliminary testing, the AI model demonstrated nearly 90% accuracy in predicting skin cancer risk across different ethnicities. While not intended to replace a doctor’s evaluation, the AI serves as a supplementary tool for earlier and more accurate identification of potential skin cancer cases.
Dr. Ben Kelley, a dermatologic surgeon at Scripps Clinic, expressed optimism about the AI's potential to bridge gaps in detection caused by infrequent or atypical cancer presentations in diverse populations. According to Kelley, AI's ability to process and identify variations in skin cancer presentation could revolutionize how dermatology clinics function, making screenings more equitable.
Future Prospects of AI in Dermatology
The success of this AI model illustrates the broader possibilities of AI in healthcare, starting with skin cancer. As more advanced models are developed, the integration of AI into clinical settings is anticipated to grow, improving early disease detection and treatment outcomes. Frazer hopes this is just the beginning of AI's role in democratizing healthcare, moving toward a future where medical assessments are both accurate and inclusive of all populations.
In summary, the work at UC San Diego highlights the transformative power of AI in medical diagnostics, offering significant promises for achieving equitable healthcare outcomes across ethnicities.
출처 : Original Source