In 1906, the English statesman Francis Galton visited a cattle fair where visitors were asked to guess the weight of an ox's suit awaiting an imminent massacre. About 800 participants took part and later Galton got the competition data.
This episode, which Galton reported in Nature, has become the object of popular repetitions, as in the 2004 book by James Surowiecki, "The Wisdom of Crowds". At 1,197 pounds, the average of all fair guess hypotheses had zero errors.
A new study by Vanderbilt University Medical Center suggests that the discovery of Galton a century ago could have implications for dermatological research and clinical evaluation.
For any number of diseases involving the skin, the search for causes and cures requires to reliably isolate and quantify the proportion of the affected skin, a research subject after the other, the more and the better it is.
This is achieved with medical photography, computer monitors and mouse tracking by a research dermatologist to delimit the affected areas with care. With the areas of interest highlighted, the software takes the final step to quantify the proportion of the affected skin.
There are massive sets of medical photographs relevant to the research, stored in hospitals and clinics, but "the time and expense of having experts questioned endlessly on these images is a major impediment, and by a study or expert at the same time. other consistency in the application of the relevant visual assessment scales tends to be poor, "said Eric Tkaczyk, MD, Ph.D., assistant professor of Dermatology and Biomedical Engineering.
Artificial intelligence is ready to provide, at a fraction of the cost, a quick and consistent automatic interpretation of such images. However, automatic learning for these evaluations will require a prodigious number of images with reliable annotations, accumulated as training sets.
"A solution to economically generate the necessary training sets could simplify research into a number of diseases and conditions and benefit patient assessment.We asked, particularly with today's economy, what results could be achieved by giving few experts pointers and allowing them to demarcate images into a Web interface. How do unedited expert assessments overlap with expert evaluation? "Tkaczyk Churches.
He and Daniel Fabbri, Ph.D., assistant professor of Biomedical Informatics, and colleagues test this notion in a new crowdsourcing study that appears in Skin Research & Technology. They tested crowd worker evaluation of a sometimes lethal chronic-versus-host disease (cGVHD).
The skin is the most commonly affected organ in cGVHD, which is the leading long-term cause of morbidity and mortality (other than cancer recurrence) after stem cell transplantation.
The study uses 41 photographs 3-D taken of patients with cGVHD. The visible weight of cGVHD in these images was first highlighted by a board-certified dermatologist with a particular interest in the disease.
Seven practitioners, in this case medical students and nurses, were given two-dimensional projections of 3D images and asked to emulate the expert, based on a slide on cGVHD and a small series of marked images. as an example guide.
The researchers evaluated the work of the pixel-by-pixel crowd. When they threw away the ends of at least most of the pixels, in terms of pixel-to-pixel correspondence with the expert evaluation, out of 410 images the median accuracy of the aggregated evaluations of four crowdfunding operators was 76%.
"This puts this group of mass workers, as a collective, very on par with the expert assessment for cGVHD," said Tkaczyk. "Our findings state that crowdsourcing could help machine learning in this area, which is beneficial for the research and clinical evaluation of this disease."
Discover a mechanism that causes a chronic, transplant-host disease after bone marrow transplantation
Eric R. Tkaczyk et al. Crowdsourcing to delineate the skin affected by a chronic graft disease against the host, Research and technology of the skin (2019). DOI: 10.1111 / srt.12688
The study of skin diseases uses crowdsourcing to collect data (2019, 22 February)
recovered on 22 February 2019
This document is subject to copyright. Regardless of any fair negotiation for private study or research, no
part can be reproduced without written permission. The content is provided for information purposes only.