TY - JOUR
T1 - A comparison of machine learning methods for the diagnosis of pigmented skin lesions
AU - Dreiseitl, Stephan
AU - Ohno-Machado, Lucila
AU - Kittler, Harald
AU - Vinterbo, Staal
AU - Billhardt, Holger
AU - Binder, Michael
N1 - Funding Information:
Science Foundation FWF, by the Max Kade Foundation (M.B.), by NLM/NHLBI Contract 467-MZ-802289 (L.O.M.), by Project Grant 107409/320 from the Norwegian Research Council (S.V.), by NLM Grant R29 LM06538-01 (L.O.M., S.V.), and by Grants FIS 97/0267 and CICYT TEL97-1073-C02-01 (HB). The authors thank H. Ganster, M.Sc., Department of Computer Graphics and Vision, Technical University Graz, for his work on the image segmentation algorithm. This work was also supported in part by Grant P-11735-MED of the FWF.
Funding Information:
Financial support for this study was provided in part by Grants J-1661-INF (S.D.) and P-11735-MED (M.B.,H.K.) from the Austrian
PY - 2001/2
Y1 - 2001/2
N2 - We analyze the discriminatory power of k-nearest neighbors, logistic regression, artificial neural networks (ANNs), decision tress, and support vector machines (SVMs) on the task of classifying pigmented skin lesions as common nevi, dysplastic nevi, or melanoma. Three different classification tasks were used as benchmarks: the dichotomous problem of distinguishing common nevi from dysplastic nevi and melanoma, the dichotomous problem of distinguishing melanoma from common and dysplastic nevi, and the trichotomous problem of correctly distinguishing all three classes. Using ROC analysis to measure the discriminatory power of the methods shows that excellent results for specific classification problems in the domain of pigmented skin lesions can be achieved with machine-learning methods. On both dichotomous and trichotomous tasks, logistic regression, ANNs, and SVMs performed on about the same level, with k-nearest neighbors and decision trees performing worse.
AB - We analyze the discriminatory power of k-nearest neighbors, logistic regression, artificial neural networks (ANNs), decision tress, and support vector machines (SVMs) on the task of classifying pigmented skin lesions as common nevi, dysplastic nevi, or melanoma. Three different classification tasks were used as benchmarks: the dichotomous problem of distinguishing common nevi from dysplastic nevi and melanoma, the dichotomous problem of distinguishing melanoma from common and dysplastic nevi, and the trichotomous problem of correctly distinguishing all three classes. Using ROC analysis to measure the discriminatory power of the methods shows that excellent results for specific classification problems in the domain of pigmented skin lesions can be achieved with machine-learning methods. On both dichotomous and trichotomous tasks, logistic regression, ANNs, and SVMs performed on about the same level, with k-nearest neighbors and decision trees performing worse.
KW - Decision support
KW - Image classification
KW - Machine learning
KW - Neural networks
KW - Support vector machines
KW - Neural Networks, Computer
KW - Diagnosis, Computer-Assisted
KW - Humans
KW - Logistic Models
KW - Melanoma/diagnosis
KW - Skin Neoplasms/diagnosis
KW - Skin Diseases/classification
KW - Algorithms
KW - Skin Pigmentation
KW - Nevus, Pigmented/diagnosis
KW - Decision Trees
KW - Nevus/diagnosis
UR - http://www.scopus.com/inward/record.url?scp=0034981810&partnerID=8YFLogxK
U2 - 10.1006/jbin.2001.1004
DO - 10.1006/jbin.2001.1004
M3 - Article
C2 - 11376540
AN - SCOPUS:0034981810
SN - 1532-0464
VL - 34
SP - 28
EP - 36
JO - Journal of Biomedical Informatics
JF - Journal of Biomedical Informatics
IS - 1
ER -