Endoscopic estimation of size: Improved accuracy by directed teaching

0016-5107/95/4204-029255.00 + .0 GASTROINTESTINAL ENDOSCOPY Copyright © 1995 by the American Society for Gastrointestinal Endoscopy

Endoscopic estimation of size: improved accuracy by directed teaching Eric Schwartz, MD, Marc F. Catalano, MD Benjamin Krevsky, MD, MPH Philadelphia, Pennsylvania Previous studies have demonstrated the inaccuracy of endoscopic estimation of size. Although several devices have been developed to help improve estimation of size, none are convenient for clinical use. We have designed and evaluated a clinical teaching protocol to aid endoscopists in better estimating size. Thirteen "endoscopists" with varying levels of experience (none, less than I year, more than I year) estimated the size of six steel ball bearings placed into a model colon and viewed with a videoendoscope. They were then taught to compensate for optic distortion and retested immediately after teaching and again I month later. The mean error of estimation decreased from 28% before teaching to 8% after teaching (p < .05) and rose to 12% 1 month later (p < .05). Although the indices of mean error decreased immediately after teaching in all groups, only those individuals with less than I year of endoscopic experience retained the improvement I month after teaching. We conclude that endoscopists can be taught how to compensate for the optic distortion encountered during endoscopy. This teaching is most effective if performed early in the training program. (Gastrointest Endosc 1995;42:292-5.)

Endoscopic estimation of size is commonly used during clinical investigations and in the evaluation of treatment efficacy.1-3 However, several previous studies have documented the inaccuracy of these clinical measurements. 4-6 Various methods have been developed to help improve endoscopic estimation of size. The simplest involves using reference markers, such as open biopsy forceps of known size. When placed into the field of view, the forceps can be used for direct comparison. However, Margnlies et al. 5 demonstrated t h a t observations made with a forceps were not significantly more accurate t h a n observations made without the forceps. The newest generation of corrective devices utilizes lasers, 6 stereoendoscopes, 7 and computer image analReceived July 19, 1994. For revision August 18, 1994. Accepted November 1, 1994. From the Temple University School of Medicine, Department of Medicine, Gastroenterology Section, Philadelphia, Pennsylvania. Reprint requests: Benjamin Krevsky, MD, MPH, Temple University Hospital, Gastroenterology, 3401 North Broad Street, Philadelphia, PA 19140. 37/1/62028 292

GASTROINTESTINAL ENDOSCOPY

ysis s to determine precisely the size of objects observed at endoscopy. Although these methods indeed improve the endoscopic estimation of size, they require special endoscopes and computer systems. They can also be very costly. The present study was designed to determine if endoscopists could be taught to compensate for optical distortion and improve endoscopic estimation of size in an easy, cost-effective manner.

MATERIALS AND METHODS Thirteen individuals were recruited to estimate the size of objects and were divided into three study groups. Group 1 consisted of five junior gastroenterology fellows with less than 1 year of endoscopic training. Group 2 consisted of four senior fellows with more than 1 year of endoscopic training. Group 3 consisted of four endoscopically untrained residents in internal medicine with no prior endoscopic experience. Objects to be sized were six steel ball bearings with an outer diameter of 3 mm, 4 mm, 7 mm, 8 mm, 12 mm, and 19 mm (Pruyn Bearing Co., Philadelphia, Pa.). These were randomly inserted for estimation into a latex model of the colon (Marks colon model, Endomodel, Inc.). Observations were made with a video colonoscope (EC-3800 series, Pentax Precision Instrument Corp., Orangeburg, N.Y.), which VOLUME 42, NO. 4, 1995

40

50

~3s

-+ 30 -'~ .=

.

45

I

40

~

.~ 25 - -

o

,~

~ 2o ~

15

--

~ o

10

--

35 30 25

o= 20 #. is L.

lO 5 o

g. 0 Initial

After teaching

F i g u r e 1. Percentage of error in underestimation of size

(mean _+ SEM) for the periods before teaching, immediately after teaching, and 1 month after teaching. The data for all three study groups are combined for each test period. *, p < .05 versus initial by ANOVA. was inserted into the model. The observer was allotted 30 seconds during which he or she could manipulate the endoscope in any fashion to estimate the bearing size to the nearest millimeter. An open biopsy forceps was placed in the field of view via the instrument channel of the endoscope to be used by the observer if desired. Ball bearings were not viewed by the subject except through the endoscopic system. After the size of all six objects had been estimated, subjects were given standardized instruction on the technique for size estimation. First, all subjects visually examined three ball bearings of known size outside the model colon without an endoscope. Second, they were instructed to place the forceps directly in contact with the ball bearings to appreciate the relative sizes of the two without endoscopic distortion. Third, the bearings of known size were placed into the model colon and the instructee was told the size. Finally, the forceps were placed into the endoscopic field of view. The subjects were instructed to touch the ball bearing and again to appreciate the relative sizes of the object and the forceps. A 2-cm graduated marker was also placed in the field for reference. This marker was present only during training sessions and was not available during testing sessions. Estimation of size of the original six ball bearings was repeated immediately after training and again approximately 1 month after training. An error index was computed, which consisted of the sum of the absolute value of the six errors of estimation (if any) for each individual, corrected for the size of the bearing. If no errors were made, the error index was zero. The formula used for the error index was: I (Actual Size--Estimated Size)/(Actual Size) I Results are presented as mean _+ standard error of the mean (SEM) and evaluated by analysis of variance (ANOVA) for repeated measures (Systat for Windows V5.03). RESULTS

T e a c h i n g r e s u l t e d in a n overall reduction in the p e r c e n t a g e of e r r o r in e s t i m a t i o n f r o m 28% to 8%. Aft e r 1 m o n t h , the e r r o r i n c r e a s e d to only 12% (Fig. 1). T h e m e a n e s t i m a t e of object size t e n d e d to be lower t h a n the a c t u a l size for e a c h b e a r i n g e v a l u a t e d . This VOLUME 42, NO. 4, 1995

Initial

1 month later

After teaching

1 month later

Figure 2. Percentage

of exactly correct estimates (mean _+ SEM) for the periods before teaching, immediately after teaching, and 1 month after teaching. The data for all three study groups are combined for each test period.

2,5

¢x •-

uJ

2 1.5 1 0.5

Initial

After teaching

1 month later

Figure 3. Error index (mean _+ SEM) for the periods before teaching, immediately after teaching, and 1 month after teaching, calculated as 7+I (actual size - estimated size)/(actual size) J. *, p < 0.05 versus initial.

w a s t r u e before a n d a f t e r endoscopic teaching, b u t the g r e a t e s t u n d e r e s t i m a t i o n s w e r e seen before t e a c h i n g (Table 1). T h e m e a n e r r o r of e s t i m a t i o n i n c r e a s e d as ballb e a r i n g size increased, b o t h before t e a c h i n g a n d a t t h e 1 - m o n t h follow-up. T h e m e a n e r r o r before t e a c h i n g w a s 0.7 m m for t h e s m a l l e s t ball b e a r i n g a n d 6.6 m m for t h e l a r g e s t ball bearing. After I m o n t h , the m e a n e r r o r for t h e s m a l l e s t ball b e a r i n g w a s 0.3 m m a n d 3.3 m m for the l a r g e s t ball bearing. I m m e d i a t e l y a f t e r training, no correlation w a s found b e t w e e n object size a n d m e a n error. E x a c t l y correct e s t i m a t e s w e r e u n c o m m o n w i t h o u t training, occurring only 14% of t h e t i m e (11 of 78 estimations). E x a c t e s t i m a t e s occurred 32% of t h e t i m e i m m e d i a t e l y a f t e r t r a i n i n g b u t fell to 24% 1 m o n t h l a t e r (Fig. 2). T h e m e a n e r r o r index (which corrected for b e a r i n g size) w a s d e t e r m i n e d for each g r o u p a n d showed a decrease i m m e d i a t e l y a f t e r t e a c h i n g for all t h r e e t e s t groups. T h e g r e a t e s t decrease in e r r o r w a s s e e n in t h e endoscopically u n t r a i n e d g r o u p (Fig. 3). Both the j u n ior fellows a n d the endoscopically u n t r a i n e d g r o u p h a d lower e r r o r indices 1 m o n t h a f t e r teaching. T h e senior fellows d e m o n s t r a t e d the lowest m e a n e r r o r GASTROINTESTINAL ENDOSCOPY

293

Table 1. Effect of teaching on endoscopic estimation of size (values are given as ±SEM) Test

Actual ball-bearing size*

period

3 mm

4 mm

7 mm

8 mm

12 m m

19 m m

Before teaching After teaching 1 m o n t h later

2.3 ± 0.4 2.5 ± 0.2 2.7 ± 0.4

2.8 ± 0.4 3.5 -+ 0.1 3.5 ± 0.3

5.1 ± 0.4 6.7 ± 0.3 6.5 ± 0.3

5,7 +_ 0.6 7,5 ± 0.5 7.0 ± 0.3

8.3 ± 0.6 11.2 ± 0.7 10.5 ± 0.4

12.4 ± 0.8 18,2 ± 1.2 15,7 ± 0.8

*Actual sizes were not k n o w n to the 13 subjects evaluated and were presented in random order.

index before teaching and improvement immediately after teaching; however, their error index returned to the preteaching level 1 month later.

DISCUSSION Endoscopic underestimation of size is in part caused by the optic distortion encountered during endoscopy. A wide-angle lens is used on an endoscope to increase the field of view. This convex lens causes a "barrel effect, "9 which makes objects in the center of a field of view appear larger than objects in the periphery. It also causes objects to appear disproportionately smaller as the distance from the endoscope increases. Numerous methods have been devised to help improve endoscopic estimation of size. The placement of various objects into the endoscopic field was among the first techniques used. Okabe et al.9 used a 5-mm disk, which was placed in the center of an ulcer model; a photograph was then taken with an overlying grid of squares of known size. From this photograph, the size of the ulcer could be estimated. They reported an average error of 5.6%. Their technique was limited by the inability to measure three-dimensional objects and also by the need to photograph objects en face. Other techniques utilized more sophisticated technology to compensate for the wide-angle lens distortion. Catalano et al.7 used a stereoendoscope to measure three-dimensional objects. They reported an error of 9.2%. Although this method was easy to learn and was reproducible, it required a unique endoscope and was very expensive, thus limiting its clinical utility. Yamaguchi et al.6 used an argon laser, a side-viewing endoscope with diffraction grating fashioned from glass fibers, and a graphic processing unit to calculate the area of disks of known size. In vitro they reported an error of 2.8%, and in vivo they found an error of 3.7%. This system, although effective in measuring three dimensions, was limited by cost and the variability seen with changes of angle of the endoscope. Recently, Vakil et al.s developed a computer program that corrects for the optical distortion seen during endoscopy. The program processes the native image by applying a correction factor to each pixel of the original image. The correction depends on the distance from the center of the image, the characteristics of the 294

GASTROINTESTINAL ENDOSCOPY

individual endoscope's optics, and other factors. They achieved an error of only 1.8% in vitro. Measurements did change with the distance of the endoscope to the object, so they used a graduated guide wire to maintain a constant distance of 3 cm. This method, although quite effective for flat lesions, cannot be used to measure three-dimensional objects. Our study demonstrates an improvement of endoscopic estimation of size after trainees are taught how to minimize the effects of optic distortion. This method emphasizes keeping the reference marker for size (in this case open biopsy forceps) close to the object. With the biopsy forceps actually touching the ball bearing, the distortion of the target object and the forceps are similar, so size can be better estimated. An error of 8% occurred after teaching, which is comparable with the error rates of the more sophisticated and costly techniques evaluated previously. All three groups demonstrated an improvement immediately after training, but only those with less than 1 year of endoscopic training demonstrated a statistically significant improvement 1 month after the teaching session. It seems that with teaching, subjects, regardless of experience, learn. Those with substantially more experience soon revert to their old habits. Further studies with continued reinforcement need to be done to see if initial improvement can be maintained for a longer period of time. In summary, we report a teaching method that minimizes the effect of endoscopic optic distortion, thus enabling more accurate estimation of the size of three-dimensional objects. This cost-effective method is comparable in its outcome with more technologically sophisticated and therefore more expensive technology. Although any individual--regardless of endoscopic experience---can be taught, those with less than 1 year of endoscopic training seem to retain the learning longer. The formal teaching of trainees how to estimate the size of objects endoscopically should be a part of the early educational process of fellowship training.

ACKNOWLEDGMENTS The latex model of the colon was provided by Pentax Precision Instrument Corporation. The authors VOLUME 42, NO. 4, 1995

would like to thank the gastroenterology fellows and medical residents of Temple University Hospital for their cooperation in the performance of this study. REFERENCES 1. Graham DY, Lew GM, Evans DG, et a]. Effects of trip]e therapy (antibiotics and bismuth) on duodenal ulcer healing. A randomized controlled trial. Ann Intern Med 1991;115:266-9. 2. Jiranek GC, Kimmey MB, Saunders DR, et al. Misoprostil reduces gastroduedenal injury from one week of aspirin: an endoscopic study. Gastroenterology 1989;96:656-61. 3. Atkin WS, Morson BC, Cuzick J. Long term risk of colorectal cancer after excision of rectosigmoid adenomas. N Engl J Med 1992;326:658-62. 4. Sonnenberg A, Giger M, Kern L, et al. How reliable is determination of ulcer size by endoscopy? BMJ 1979;2:1322-4.

VOLUME 42, NO. 4, 1995

5. Margulies C, Krevsky B, Catalano MF. How accurate are endoscopic estimates of size? Gastrointest Endosc 1993;40: 174-7. 6. Yamaguchi M, Okazaki Y, Yanal H, et al. Three dimensional determination of gastric ulcer size with laser endoscopy. Endoscopy 1988;20:263-6. 7. Catalano MF, Van Dam J, Bedford R, et al. Preliminary evaluation of the prototype stereoendoscopic endoscope: precise three-dimensional measurement system. Gastrointest Endosc 1993;39:23-8. 8. Vakil N, Smith W, Bourgeois K, et al. Endoscopic measurement of lesion size: improved accuracy with image processing. Gastrointest Endosc 1994;40:178-83. 9. Okabe H, Ohida M, Okada N, et al. A new disk method for endoscopic determination of gastric ulcer area. Gastrointest Endosc 1986;32:20-4.

G A S T R O I N T E S T I N A L E N D O S C O P Y 295