 |
|
Variability in Physician Assessment of Lesions in Cutaneous Images and Its Implications for Skin Screening and Computer-Assisted Diagnosis
Douglas A. Perednia, MD;
John A. Gaines, PhD;
Alfred C. Rossum, MD
Arch Dermatol. 1992;128(3):357-364.
Abstract
| |
• Background and Design.—
The diagnosis of skin lesions is a three-part process of (1) screening for abnormalities, (2) examining lesions closely for the presence of important clinical characteristics, and (3) choosing a diagnosis based on the clinical data gathered. The first step, screening for abnormal areas, is of particular interest for anyone wishing to understand and automate the diagnostic process. As part of ongoing research into lesion detection and measurement, we examined the reproducibility and reliability of dermatologists who were asked to screen sets of cutaneous digital images and high-resolution photographs for the presence of skin lesions.
Results.—
The intraobserver and interobserver variation in this task was considerable. The average reproducibilities and reliabilities were approximately 85%. The photographs did not produce significantly better overall screening results than the moderately low-resolution digital images.
Conclusions.—
Even highly trained dermatologists appear to differ significantly in their initial interpretations of skin images, both from observer to observer and within observers from session to session. These differences were reflected in all measures of diagnostic sensitivity, reproducibility, and reliability assessed. The magnitude of this variation is relatively insensitive to the means used to measure it, and is not simply due to different individual standards for diagnosis. It is highly unlikely that this variation is attributable to the images themselves, since the overall results were nearly identical for moderately low-resolution digital formats and very high-resolution photographs. These findings have important implications for determining where errors may take place in dermatologic screening and for automating the process of detecting changes in skin lesions over time.
(Arch Dermatol. 1992;128:357-364)
Author Affiliations
From the Department of Dermatology, Oregon Health Sciences University, Portland (Dr Perednia), the Office of Biostatistical Services, University of Arizona Health Sciences Center, Tucson (Dr Gaines), and the Department of Medicine, University of New Mexico Hospital, Albuquerque (Dr Rossum).
Footnotes
Accepted for publication November 7, 1991.
Reprint requests to Department of Dermatology, Oregon Health Sciences University, 3181 SW Sam Jackson Park Rd, L468, Portland, OR 97201-3098 (Dr Perednia).
 CiteULike  Connotea  Del.icio.us  Digg  Reddit  Technorati  Twitter
What's this?
THIS ARTICLE HAS BEEN CITED BY OTHER ARTICLES
Development of clinical sign based algorithms for community based assessment of omphalitis
Mullany et al.
Arch. Dis. Child. Fetal Neonatal Ed. 2006;91:F99-F104.
ABSTRACT
| FULL TEXT
Do We Have Time for the Change?
Margolis
Arch Dermatol 1998;134:1151-1152.
FULL TEXT
Teledermatology and In-Person Examinations: A Comparison of Patient and Physician Perceptions and Diagnostic Agreement
Lowitt et al.
Arch Dermatol 1998;134:471-476.
ABSTRACT
| FULL TEXT
Comparison of the Clinical Informativeness of Photographs and Digital Imaging Media With Multiple-Choice Receiver Operating Characteristic Analysis
Perednia et al.
Arch Dermatol 1995;131:292-297.
ABSTRACT
The Measure of Youth
Stern and Coopman
Arch Dermatol 1992;128:390-393.
ABSTRACT
|
