Error in estimates of size

Abstract

Two recent articles describing size underestimation of objects viewed endoscopically1Marguiles C Krevsky B Catalano M. How accurate are endoscopic estimates of size?.Gastrointest Endosc. 1994; 40: 174-177Abstract Full Text Full Text PDF PubMed Scopus (83) Google Scholar, 2Vakil N Smith W Bourgeois K Everbach EC Knyrim K. Endoscopic measurement of lesion size: improved accuracy with image processing.Gastrointest Endosc. 1994; 40: 178-183Abstract Full Text Full Text PDF PubMed Scopus (57) Google Scholar and an abstract demonstrating improvement in size estimates after training3Schwartz E Krevsky B. Endoscopic estimation of size: can it be taught?.Gastrointest Endosc. 1994; 40 ([Abstract]): 47Google Scholar prompted simple size estimate experiments independent of endoscopy. These experiments show underestimation of size may be independent of endoscopic view and inherent distortion and a function of human error. Six examiners (three gastroenterology fellows, three gastroenterology faculty) were given a one cent coin (a penny) and asked to estimate its diameter in millimeters. They were allowed to handle the coin, but not any measuring device (measurements improve size estimates but do not eliminate error). Five of six examiners underestimated the size by an average of over 28% (four of these five were wearing corrective lenses). One (also wearing corrective lenses) overestimated the size by over 5%. This is remarkably similar to the 3% overestimate to 33% underestimate noted in the reported error in endoscopic estimates.1Marguiles C Krevsky B Catalano M. How accurate are endoscopic estimates of size?.Gastrointest Endosc. 1994; 40: 174-177Abstract Full Text Full Text PDF PubMed Scopus (83) Google Scholar Five of six examiners wore corrective lenses. Examiners' ages ranged from 30 to 63 years, with the average age being 42 years. Age or corrective lens use did not appear to confer a significant advantage, although the sample size is quite small ( Table 1 ).Table 1Estimates and errorsObserverEstimate of penny (mm)Error (%)ErrorEstimate of quarter (mm)Error (%)ErrorError index*Age (yr)Corr. lens.115-21.05-4240.00021.0533Yes2205.2613545.831151.1046Yes316-15.79-32816.67432.4644Yes411-42.11-8254.17146.2736No513-31.58-62816.67448.2530Yes613-31.58-6240.00031.5863YesAverage14.67-22.81-4.3327.3313.893.3338.4542Actual1924SD3.1416.533.144.1817.414.1811.86*Sum of absolute per cent error. Open table in a new tab A test of whether brief training could improve size estimates was then conducted. The measured size of the penny (approximately 19 mm) together with individual (range 13 to 20 mm) and average (14.67 mm) results of the previous size experiment were immediately revealed and discussed in connection with the cited articles. This discussion served as a “training session” in estimating size. Thereafter, the same examiners were asked to estimate the size of a twenty-five cent coin (a quarter, measured as approximately 24 mm). Results of this experiment ( Table 1 ) show improvement in size estimates to an average overestimate of about 14% (a reduction in error of (23-14)/23 or 39%). In addition, although no examiner estimated the size of the penny “exactly,” after training two of six estimated the size of the quarter exactly. This level of improvement correlates with the results following training cited in the abstract.3Schwartz E Krevsky B. Endoscopic estimation of size: can it be taught?.Gastrointest Endosc. 1994; 40 ([Abstract]): 47Google Scholar Again, age or corrective lens use did not appear to be a significant factor, with the two examiners reaching an exact estimate of the quarter being toward the extreme ends of age distribution. Endoscopic estimates of size, with or without comparison to objects in the visual field, are subject to gross error. Estimates of size are susceptible to gross human error independent of endoscopy, yet can be improved with specific training and reinforcement. The simple experiments outlined here may be used to demonstrate human error in estimates of size, provide training, and reinforce correct estimates of size. Improvement in estimates of size of common objects that can be held and manipulated may extend to improvement in estimates of size of images seen endoscopically. Others are invited to include other common small objects to reproduce these results and achieve improvement in size estimates.