| Abstract |
Accurate diagnosis and classification of epilepsies and epileptic syndromes allow the scientific analysis of the underlying disease processes and their specific clinicopathological features and genetics and provides a framework for clinical trials aimed at optimising treatment. Because of their atypical appearance their polymorphism, diagnosis and differential diagnosis of epilepsies in childhood is often difficult. A computerized Medical Decision Support System (MDSS) could be helpful for the classification and diagnosis of Epilepsy. Methods from Artificial Intelligence have supported a wide variety of clinical and medical decision making tasks. Furthermore, Artificial Intelligence methods have contributed towards the formalization and representation of medical knowledge both formal and informal. Formal connotes knowledge embodied in textbooks and well-established procedures while informal indicates context and subjective elements. Alternative methods inspired from Artificial Intelligence have supported a wide variety of clinical and medical decision-making tasks. Since pure Artificial Intelligence methods offer mainly a clustering of the data, decision support systems could be more appropriate to support diagnosis in that they are information-processing systems and are particularly helpful when quick and correct decision-making is needed. Such systems support the clinician and help trainees/generalists to perform quicker and more accurate diagnosis. The dissertation reports research process and results underlying the development and assessment of a medical decision aid, which aims to support medical doctors in the diagnosis of epilepsy with special emphasis on childhood episodes. Classification of epileptic syndromes is done according to the 'International Classification of Epileptic Syndromes and Epilepsies' (Commission on Classification and Terminology of the ILAE 1981, 1989) set forth by the as is proposed by the 'International League against Epilepsy' (ILAE). ILAE nomenclature, procedure and standards are extensively used in modeling presented herein. International Classification includes more than 50 diagnostic categories for epilepsy and epileptic syndromes. To account the diagnostic categories more than 100 different factors (lab and EEG findings, symptoms, clinical data, etc) affecting the diagnosis should be assessed and effectively incorporated in a decision aid system. Methodology decomposes diagnosis into smaller size [sub] diagnosis instances. Thus, size and complexity are reduced with no loss in diagnostic accuracy and clinical comprehensibility of result. Inference draws on decision trees, which are automatically induced using specific input data. We selected to work with decision trees because they provide a convenient tool, which also conforms to clinical decision-making and differential diagnosis. The system and the physician reached identical diagnoses in 85.2% of the cases. In an additional 8.2% of the cases, the system's diagnosis was similar to that of the physician thus raising its overall success rate to 93.4%. The system can be helpful especially for trainees since it only needs to import the clinical and laboratory data. Decision-making and differential diagnosis is then performed automatically. Our preliminary results encourage us to support the use of this decision support system. In general, the diagnostic system we constructed led to the right diagnosis for the majority of the cases. It is certain that no matter how accurate and rational the diagnostic ability of the experienced system may be, it is always a result of series of finite data analyses, so the treating doctor should always have the last word. Nevertheless, he can always refer to it, in order to verify the diagnosis. The software's results can be of consultative character. A very efficient role of the specific diagnostic system is certainly that of an educational tool for students of medicine or doctors in the period of their specialization, as its development has been based on the international classification of syndromes and seizures.
|
Ανάπτυξη συστήματος για την διάγνωση επιληψιών και επιληπτικών σύνδρομων σε παιδιά με τη βοήθεια υπολογιστών
