ΕΛΛΗΝΙΚΟ ΜΕΣΟΓΕΙΑΚΟ ΠΑΝΕΠΙΣΤΗΜΙΟ
ΤΜΗΜΑ ΗΛΕΚΤΡΟΛΟΓΩΝ ΜΗΧΑΝΙΚΩΝ ΚΑΙ ΜΗΧΑΝΙΚΩΝ ΥΠΟΛΟΓΙΣΤΩΝ
ΠΡΟΣΚΛΗΣΗ ΣΕ ΔΗΜΟΣΙΑ ΠΑΡΟΥΣΙΑΣΗ
ΜΕΤΑΠΤΥΧΙΑΚΗΣ ΔΙΠΛΩΜΑΤΙΚΗΣ ΕΡΓΑΣΙΑΣ
Τίτλος
«A Web-Based Differential Diagnosis System based on ICD10 using multiple knowledge bases»
του Άρη Παπακωνσταντίνου, μεταπτυχιακού φοιτητή του
Προγράμματος Μεταπτυχιακών Σπουδών “Μηχανικών Πληροφορικής”
Επιβλέπων: Εμμανουήλ Μαρακάκης, Ομότιμος Καθηγητής
H παρουσίαση θα πραγματοποιηθεί την Τετάρτη 22 Ιουλίου 2026 στις 9:00 πμ διαδικτυακά μέσω του συνδέσμου https://vdc.hmu.gr/b/mar-fto-wlp-vyv.
ΠΕΡΙΛΗΨΗ ΤΗΣ ΕΡΓΑΣΙΑΣ
This thesis presents an innovative web-based differential diagnosis system, aiming to enhance diagnostic precision. The proposed system utilizes Prolog-based knowledge representation and meta-rules to analyze patient symptoms and generate differential diagnoses with confidence scores. It features two distinct diagnostic engines operating in parallel: one prioritizes symptom pattern recognition from established medical textbooks, while the other employs pathophysiological reasoning to identify underlying disease origins.
The system’s architecture consists of three main layers: a web interface layer for user interaction, a diagnostic subsystem layer that manages processing and logical analysis, and a knowledge base layer containing medical knowledge derived from established sources. The integration of ICD-10 codes supports compatibility with existing healthcare systems and standardizes diagnostic outputs.
The system was evaluated using a dataset of 50 respiratory cases with expert-confirmed diagnoses. Results indicate that the first diagnostic sub-system achieved 55% accuracy, the second achieved 30% accuracy, and the integrated system achieved 60% accuracy. The system showed particularly strong performance in common respiratory conditions, with 92% accuracy in acute respiratory cases.
The contributions of this work include: (1) the development of a dual knowledge base architecture that combines different diagnostic methodologies, (2) the integration of ICD-10 codes for standardized outputs, (3) the generation of transparent explanations for diagnostic recommendations, and (4) the creation of a user-friendly web interface that facilitates clinical decision-making.
Future work will prioritize extending the knowledge bases to cover additional medical specialties, incorporating machine learning techniques for improved accuracy, and conducting larger-scale clinical evaluations to confirm the system’s effectiveness in real-world healthcare environments.