Machine classification of melanoma and nevi from skin lesions

Authors:
John David Osborne;Song Gao;Wei-bang Chen;Aleodor Andea;Chengcui Zhang
Affiliations:
The University of Alabama at Birmingham, Birmingham, Alabama;The University of Alabama at Birmingham, Birmingham, Alabama;The University of Alabama at Birmingham, Birmingham, Alabama;The University of Alabama at Birmingham, Birmingham, Alabama;The University of Alabama at Birmingham, Birmingham, Alabama
Venue:
Proceedings of the 2011 ACM Symposium on Applied Computing
Year:
2011

Citing 3
Cited 1

Computer-aided evaluation of neuroblastoma on whole-slide histology images: Classifying grade of neuroblastic differentiation

Pattern Recognition
Segmentation of blood and bone marrow cell images via learning by sampling

ICIC'09 Proceedings of the 5th international conference on Emerging intelligent computing technology and applications
LIBSVM: A library for support vector machines

ACM Transactions on Intelligent Systems and Technology (TIST)

Color Image Segmentation: From the View of Projective Clustering

International Journal of Multimedia Data Engineering & Management

Quantified Score

Hi-index	0.00

Visualization

Abstract

We describe a method using a Support Vector Machine (SVM) to classify and diagnose skin biopsies from patients as either melanoma or nevi based on H&E stained histological slides alone. Our method differs from other approaches to digital melanoma diagnoses in using the histology slide, not digital clinical pictures of the patients' skin to make the classification. Using only the histological criterion of irregularities in the nucleus, our best SVM utilizes nucleus perimeter/area ratio and nucleus major/minor axis ratio as features to give a classification accuracy of 90%, sensitivity of 100% and specificity of 75%, (at magnification of 400 times) in our data set. The performance is remarkable given a dermatological pathologist typically examines a plethora of features to make a diagnosis. Our SVM in conjunction with clinical digital diagnoses systems could reduce the number of missed melanoma diagnoses.