Author Interviews, Cancer Research, Dermatology, Lancet, Melanoma, Technology / 11.11.2021
Dermatology: Datasets Used for AI Lack Diversity and Completeness
MedicalResearch.com Interview with:
[caption id="attachment_58371" align="alignleft" width="150"]
Dr. Wen[/caption]
Dr David Wen BM BCh
NIHR Academic Clinical Fellow in Dermatology
University of Oxford
MedicalResearch.com: What is the background for this study?
Response: Publicly available skin image datasets are commonly used to develop machine learning (ML) algorithms for skin cancer diagnosis. These datasets are often utilised as they circumvent many of the barriers associated with large scale skin lesion image acquisition. Furthermore, publicly available datasets can be used as a benchmark for direct comparison of algorithm performance.
Dataset and image metadata provide information about the disease and population upon which the algorithm was trained or validated on. This is important to know because machine learning algorithms heavily depend on the data used to train them; algorithms used for skin lesion classification frequently underperform when tested on independent datasets to which they were trained on. Detailing dataset composition is essential for extrapolating assumptions of generalisability of algorithm performance to other populations.
At the time this review was conducted, the total number of publicly available datasets globally and their respective content had not previously been characterised. Therefore, we aimed to identify publicly available skin image datasets used to develop ML algorithms for skin cancer diagnosis, to categorise their data access requirements, and to systematically evaluate their characteristics including associated metadata.
Dr. Wen[/caption]
Dr David Wen BM BCh
NIHR Academic Clinical Fellow in Dermatology
University of Oxford
MedicalResearch.com: What is the background for this study?
Response: Publicly available skin image datasets are commonly used to develop machine learning (ML) algorithms for skin cancer diagnosis. These datasets are often utilised as they circumvent many of the barriers associated with large scale skin lesion image acquisition. Furthermore, publicly available datasets can be used as a benchmark for direct comparison of algorithm performance.
Dataset and image metadata provide information about the disease and population upon which the algorithm was trained or validated on. This is important to know because machine learning algorithms heavily depend on the data used to train them; algorithms used for skin lesion classification frequently underperform when tested on independent datasets to which they were trained on. Detailing dataset composition is essential for extrapolating assumptions of generalisability of algorithm performance to other populations.
At the time this review was conducted, the total number of publicly available datasets globally and their respective content had not previously been characterised. Therefore, we aimed to identify publicly available skin image datasets used to develop ML algorithms for skin cancer diagnosis, to categorise their data access requirements, and to systematically evaluate their characteristics including associated metadata.
Kelly Gavigan[/caption]
Kelly Gavigan, MPH
Manager, Research and Data Science
Dr. Elwy[/caption]
Rani Elwy, PhD
Bridge Quality Enhancement Research Initiative Program, Center for Healthcare Organization and Implementation Research,
VA Bedford Healthcare System
Bedford, Massachusetts
Department of Psychiatry and Human Behavior, Alpert Medical School
Brown University, Providence, Rhode Island
MedicalResearch.com: What is the background for this study?
Response: The VA operates a very robust, embedded quality improvement and implementation science program, of which our team is involved. As the VA was one of the first US healthcare systems to rollout COVID-19 vaccination programs, we were asked to evaluate these efforts in real-time, to provide input to VA healthcare leaders on what was going well and what could be improved. This survey reported in JAMA Network Open is one of the quality improvement efforts we engaged in.
Dr. Hunter[/caption]
Jennifer Hunter, B.Med., M.Sc.P.H., Ph.D.
Adjunct Associate Professor
NICM Health Research Institute
Western Sydney University
Associate Professor Jennifer Hunter is an academic general practitioner with a clinical interest in integrative medicine, has received payment for providing expert advice about traditional, complementary and integrative medicine, including nutraceuticals, to industry, government bodies and non-government organisations, and spoken at workshops, seminars and conferences for which registration, travel and/or accommodation has been paid for by the organisers.
MedicalResearch.com: What is the background for this study?
Response: We decided to review the evidence for zinc in response to calls for rapid evidence reviews to inform self-care and clinical practice during the COVID-19 pandemic.
Laboratory studies have found that zinc can inhibit the replication of many respiratory viruses, including SARS-CoV-2 and other coronaviruses. Zinc plays a key role in immunity, inflammation, tissue injury, ACE-2 receptor activity, and also in tissue responses to a lack of oxygen. Low zinc status may be a risk factor for severe SARS-CoV-2 illness.
Additionally, there was some indirect evidence suggesting zinc might be effective for other respiratory tract infections such as the common cold and we wanted to verify this.
Dr. Howard[/caption]
Jeffrey Howard, PhD
Associate Professor
Department of Public Health
College for Health, Community and Policy
University of Texas at San Antonio
MedicalResearch.com: What is the background for this study?
Response: Drug and alcohol related mortality has been on the rise in the US for the past decade, which has drawn a lot of focus from researchers. At the same time maternal mortality, deaths caused by pregnancy complications, is recognized to be higher in the US than in other developed nations.
Very little has been reported about deaths among pregnant and recently pregnant women that are not caused by pregnancy complications, so my collaborators and I wanted to explore this. We did not anticipate that drug and alcohol deaths and homicides would account for so many deaths among pregnant and recently pregnant women.
Dr. Dan P. Ly[/caption]
Division of General Internal Medicine and Health Services Research
David Geffen School of Medicine at UCLA
Los Angeles, CA
MedicalResearch.com: What is the background for this study?
Response: Lyme disease presents first on the skin with the classic “bull’s-eye” rash. But such rashes in Black patients aren’t well-represented in medical textbooks. This may lead to physicians not recognizing such rashes in Black patients.
As a result, Black patients are more likely to present with later complications of Lyme disease when first diagnosed such as neurologic complications.