ABSTRACT
BACKGROUND: Chronic graft-versus-host disease (cGVHD) is a significant cause of long-term morbidity and mortality in patients after allogeneic hematopoietic cell transplantation. Skin is the most commonly affected organ, and visual assessment of cGVHD can have low reliability. Crowdsourcing data from nonexpert participants has been used for numerous medical applications, including image labeling and segmentation tasks. OBJECTIVE: This study aimed to assess the ability of crowds of nonexpert raters-individuals without any prior training for identifying or marking cGHVD-to demarcate photos of cGVHD-affected skin. We also studied the effect of training and feedback on crowd performance. METHODS: Using a Canfield Vectra H1 3D camera, 360 photographs of the skin of 36 patients with cGVHD were taken. Ground truth demarcations were provided in 3D by a trained expert and reviewed by a board-certified dermatologist. In total, 3000 2D images (projections from various angles) were created for crowd demarcation through the DiagnosUs mobile app. Raters were split into high and low feedback groups. The performances of 4 different crowds of nonexperts were analyzed, including 17 raters per image for the low and high feedback groups, 32-35 raters per image for the low feedback group, and the top 5 performers for each image from the low feedback group. RESULTS: Across 8 demarcation competitions, 130 raters were recruited to the high feedback group and 161 to the low feedback group. This resulted in a total of 54,887 individual demarcations from the high feedback group and 78,967 from the low feedback group. The nonexpert crowds achieved good overall performance for segmenting cGVHD-affected skin with minimal training, achieving a median surface area error of less than 12% of skin pixels for all crowds in both the high and low feedback groups. The low feedback crowds performed slightly poorer than the high feedback crowd, even when a larger crowd was used. Tracking the 5 most reliable raters from the low feedback group for each image recovered a performance similar to that of the high feedback crowd. Higher variability between raters for a given image was not found to correlate with lower performance of the crowd consensus demarcation and cannot therefore be used as a measure of reliability. No significant learning was observed during the task as more photos and feedback were seen. CONCLUSIONS: Crowds of nonexpert raters can demarcate cGVHD images with good overall performance. Tracking the top 5 most reliable raters provided optimal results, obtaining the best performance with the lowest number of expert demarcations required for adequate training. However, the agreement amongst individual nonexperts does not help predict whether the crowd has provided an accurate result. Future work should explore the performance of crowdsourcing in standard clinical photos and further methods to estimate the reliability of consensus demarcations.
ABSTRACT
BACKGROUND: Artificial intelligence (AI) and machine learning (ML) technology design and development continues to be rapid, despite major limitations in its current form as a practice and discipline to address all sociohumanitarian issues and complexities. From these limitations emerges an imperative to strengthen AI and ML literacy in underserved communities and build a more diverse AI and ML design and development workforce engaged in health research. OBJECTIVE: AI and ML has the potential to account for and assess a variety of factors that contribute to health and disease and to improve prevention, diagnosis, and therapy. Here, we describe recent activities within the Artificial Intelligence/Machine Learning Consortium to Advance Health Equity and Researcher Diversity (AIM-AHEAD) Ethics and Equity Workgroup (EEWG) that led to the development of deliverables that will help put ethics and fairness at the forefront of AI and ML applications to build equity in biomedical research, education, and health care. METHODS: The AIM-AHEAD EEWG was created in 2021 with 3 cochairs and 51 members in year 1 and 2 cochairs and ~40 members in year 2. Members in both years included AIM-AHEAD principal investigators, coinvestigators, leadership fellows, and research fellows. The EEWG used a modified Delphi approach using polling, ranking, and other exercises to facilitate discussions around tangible steps, key terms, and definitions needed to ensure that ethics and fairness are at the forefront of AI and ML applications to build equity in biomedical research, education, and health care. RESULTS: The EEWG developed a set of ethics and equity principles, a glossary, and an interview guide. The ethics and equity principles comprise 5 core principles, each with subparts, which articulate best practices for working with stakeholders from historically and presently underrepresented communities. The glossary contains 12 terms and definitions, with particular emphasis on optimal development, refinement, and implementation of AI and ML in health equity research. To accompany the glossary, the EEWG developed a concept relationship diagram that describes the logical flow of and relationship between the definitional concepts. Lastly, the interview guide provides questions that can be used or adapted to garner stakeholder and community perspectives on the principles and glossary. CONCLUSIONS: Ongoing engagement is needed around our principles and glossary to identify and predict potential limitations in their uses in AI and ML research settings, especially for institutions with limited resources. This requires time, careful consideration, and honest discussions around what classifies an engagement incentive as meaningful to support and sustain their full engagement. By slowing down to meet historically and presently underresourced institutions and communities where they are and where they are capable of engaging and competing, there is higher potential to achieve needed diversity, ethics, and equity in AI and ML implementation in health research.