Support Center Machine Method for Classification and Help in Medical Diagnosis SystemZefeng Wang1*, Laurent Peyrodie2, Hua Cao2 and Samuel Boudet2
- *Corresponding Author:
- Zefeng Wang
Associate professor, Catholic University of Lille
FMM (HEI-USTB), 41 rue du port, Lille
Nord 59000, France
E-mail: [email protected]; [email protected]
Received date: July 06, 2015; Accepted date: July 30, 2015; Published date: August 10, 2015
Citation: Wang Z, Peyrodie L, Cao H, Boudet S (2015) Support Center Machine Method for Classification and Help in Medical Diagnosis System. J Theor Comput Sci 2:129. doi:10.4172/2376-130X.1000129
Copyright: © 2015 Wang Z, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Objectives: A new artificial intelligent method ‘Support Center Machines’ (SCM) for helping diagnosis and prognosis is applied to a medical system. Methods In data processing, SCM seeks the true centers of each class during machine learning. For application in the medical system, it makes these centers as health-situation models and translates all the health records into a map. All the models, like non disease and diseases, are labeled in this map. Thus, the evolution of patient’s health records can be supervised with the map. On the basis of the evolution of the distances from the recent record data to the centers, the system estimates the tendency of the healthy evolution and forecasts the probable situation in the future. Results: SCM was tested on ‘Wisconsin Breast Cancer Data’ and compared with LDA and SVM methods. Twenty centers were found to define the healthy map. Based on the test results of four hundred and fifty random data selection for train, SCM has shown a better performance, whose means of correct detection ratios of the breast cancer varied from 91.4% to 95.6%, which were corresponding to 10% of data and 90% of data used to do machine learning. These ratios have increased by 1% to 5%, than SVM and LDA. In addition, the variance of correct detection ratios of SCM results has decreased by 0.8% to 3.0%, compared with SVM and LDA. Even if there were only 10% data for the training, the ratio stayed around 87% with only 3 principal components. When the system used 50% data for the training and tests the others, the mean of ratio was 93% and the best was 95%.
Conclusions: SCM successfully build a disease diagnosis/prognosis system and works out a healthy map. It could display the health record on 2D or 3D map, which allows the clinician to appropriate the interpretation. In addition, if a new situation (symptom / disease) occurs, the practitioner can visualize it and analyze it according to existing maps of SCM.