1. MODEL SYSTEMU SAMOUCZĄCEGO W DIAGNOSTYCE MEDYCZNEJ.
- Author
-
Giżewski, Tomasz, Kowalski, Ireneusz M., Zarzycki, Daniel, Radomska-Wilczewska, Anna, Lewandowski, Roman, and Kotwicki, Tomasz
- Subjects
- *
ARTIFICIAL neural networks , *DIAGNOSTIC imaging , *ORTHOPEDIC diagnosis , *MEDICAL rehabilitation , *SCOLIOSIS , *SPECTRUM analysis , *MEDICAL imaging systems - Abstract
Introduction. The era of computer testing, measurement and decision-making systems began in the second half of the 20th century. Medical application of such systems manifests itself in digital devices for the observation of the patient's condition, X-ray and laboratory diagnostics. In this article the authors concentrate on the decision-making systems, and, in particular, on decision-making self-learning systems which, as the authors suggest, could be applied in medical diagnostics. Creating the model of the self-learning system based on patients' clinical records for the use of rehabilitation-orthopedic diagnostics is the principal aim of this research. Materials and methods. Pathological entities displayed by 850 patients (at the age of 6-20 years, on average 14.6 years) were classified. All the patients underwent treatment for idiopathic scoliosis (IS) in the Department and Clinic of Rehabilitation and in the Provincial Rehabilitation Hospital for Children in Ameryka near Olsztyn, Poland. The diagnostics imaging technique (X-ray, CT, CT-3D) was focused on identifying characteristic entities, discovered by the image algorithm identification and by the analysis of spectrum obtained out of Fourier transformation and Falk transformation. As regards the IS identification, it was appropriate to use diagnostics imaging techniques, as well as anamnesis to analyze the data. Results and discussion. The research resulted in creating the output vector in the form of anticipated treatment effects. The topology of the classification network was designed on the basis of the artificial neural networks recurrent algorithm. Conclusions. The applied classification procedure proves that defining input data and normalizing reply vector allows researchers to build the artificial neural network for the purpose of anticipating treatment effects on the basis of the acquired knowledge. For a competent medical doctor, the self-learning system can function only as an advisory device, which produces a useful information tool by means of accumulated information. [ABSTRACT FROM AUTHOR]
- Published
- 2008