Henning JS, Stein JA, Yeung J, Dusza SW, Marghoob AA, Rabinovitz HS, Polsky D, Kopf AW. CASH Algorithm for Dermoscopy Revisited. Arch Dermatol. 2008;144(4):554-555. doi:10.1001/archderm.144.4.554
Our group recently described a new dermoscopic algorithm, CASH (color, architecture, symmetry, homogeneity) to evaluate melanocytic neoplasms.1 Herein we compare CASH to 3 other algorithms: ABCD,2 the method used by Menzies et al (hereinafter, the Menzies mehtod),3 and the 7-point checklist.4
For this study, photographs of 150 melanocytic neoplasms (50 malignant melanomas, 50 dysplastic nevi, and 50 common nevi) were selected from 1535 images from the American Academy of Dermatology DVD consortium and the private collection of 1 of us (A.W.K.). Twenty-one of the melanomas were in situ. Each lesion had a clinical and dermoscopic image. Final diagnoses were based on histopathologic examination after informed consent was obtained. The original magnification of the images was ×10 in all cases. They were taken either digitally (DermLite; 3Gen LLC, San Juan Capistrano, California) or as 35-mm color transparencies (Heine Dermaphot; Heine Optotechnik, Herrsching, Germany) and scanned as JPEG images. Lesions with poor image quality were excluded. The first 50 good-quality images in each category were included. These lesions were different from the ones in our group's prior study.1
The evaluators were 2 dermatology residents with less than 1.2 years of dermoscopic experience who had not participated in our group's prior study.1 One had taken two 1-day dermoscopy courses, and the other had spent 2 months studying with Scott W. Menzies, MD. They had no other specific dermoscopic training.
The evaluators independently reviewed the paired clinical and dermoscopic images of the selected 150 lesions using 4 different dermoscopic algorithms in no particular order: CASH,1 ABCD,2 the Menzies method,3 and the 7-point checklist.4 Details of these algorithms can be found elsewhere.1- 4 Both investigators were blinded to the diagnoses.
The outcome variable in this study was dichotomous (benign melanocytic nevus or malignant melanoma) for each study lesion. The sensitivities and specificities were calculated for each algorithm and were compared with those of CASH.
The sensitivities of all 4 algorithms ranged from 76% for the 7-point checklist to 92% for the Menzies method (Table). None showed a statistically significant difference compared with CASH. The sensitivity of CASH was similar to that of ABCD (87% vs 86%). However, CASH showed a significantly higher specificity than the Menzies method and the 7-point checklist. The CASH and ABCD algorithms did not have statistically significant different specificities.
As CASH and ABCD have higher specificities than the other two algorithms, they would be less likely to result in unnecessary biopsies.
Both CASH and ABCD use very similar criteria, so it is not surprising that the 2 algorithms performed similarly in our study. Both evaluate color, symmetry, and dermoscopic structures. The new feature that CASH introduces is architecture, which forces the user to make a judgment about the overall organization of the lesion. This is an important skill used in the more sophisticated, though difficult-to-learn, technique of pattern analysis, a technique used by expert dermoscopists. The CASH algorithm provides a regimented way of teaching pattern analysis to the novice dermoscopist. Another benefit of CASH is that it does not require any weighting factors to calculate a total score, making it quicker and easier to use.
We hope to repeat this study with a large number of evaluators to further validate the CASH algorithm in a consensus Internet meeting on dermoscopy.5
Correspondence: Dr Stein, New York University School of Medicine, Ronald O. Perelman Department of Dermatology, 550 First Ave, New York, NY 10016 (firstname.lastname@example.org).
Financial Disclosure: None reported.