Neural networks assisted diagnosis of ischemic CVA's through CT scan

Technological and computing evolution promoted new opportunities to improve the quality of life through new medical achievements, in particular, the quality of diagnostic evaluations. Computerised tomography (CT) is one of the imaging equipments for diagnosis which has most benefited from technological improvements. Because of that, and due to the quality of the diagnosis produced, it is one of the most employed equipments in clinical applications. As an example, the ischaemic cerebral vascular accident (ICVA) is a pathology confirming the frequent use of CT. The interest in this pathology, and in general for the encephalon image analysis as a preventive diagnosis, is mainly due to its frequent occurrence in development countries and its social- economic impact. In this paper we propose to evaluate the ability of artificial neural networks (ANNs) for automatic identification of ICVAs by means of tissue density images obtained by CT. Cranioencephalon CT exams and their respective medical reports were used to train ANN classifiers by means of features extracted from the images. Once the ANNs were trained, the classifiers were tested with data never seen by the network. At this stage we may conclude that the ANNs may significantly contribute as an ICVAs CT diagnostic aid, since among the test cases the automatic identification of ischaemic lesions has been performed with no false negatives and very few false positives.

[1]  D. Marquardt.  An Algorithm for Least-Squares Estimation of Nonlinear Parameters , 1963 .

[2]  António E. Ruano,et al.  Solar radiation prediction using RBF Neural Networks and cloudiness indices , 2006, The 2006 IEEE International Joint Conference on Neural Network Proceedings.

[3]  J. Stephen Judd,et al.  Learning in neural networks , 1988, COLT '88.

[4]  Rafael Wiemker,et al.  Computer-aided detection (CAD) and volumetry of pulmonary nodules on high-resolution CT data , 2003 .

[5]  W. L. Robb,et al.  Perspective on the first 10 years of the CT scanner industry. , 2003, Academic radiology.

[6]  J. Champaign,et al.  Advances in breast cancer detection with screening mammography. , 2000, The Ochsner journal.

[7]  António E. Ruano,et al.  Exploiting the separability of linear and nonlinear parameters in radial basis function networks , 2000, Proceedings of the IEEE 2000 Adaptive Systems for Signal Processing, Communications, and Control Symposium (Cat. No.00EX373).

[8]  Carlos M. Fonseca,et al.  An overview of nonlinear identification and control with neural networks. , 2005 .

[9]  J M Wardlaw,et al.  RADIOLOGY OF STROKE , 2001, Journal of neurology, neurosurgery, and psychiatry.

[10]  Carlo H. Séquin,et al.  Optimal adaptive k-means algorithm with dynamic adjustment of learning rate , 1995, IEEE Trans. Neural Networks.