The Faculty of Computer Science and Mathematics discussed a doctoral dissertation entitled “Heart Disease Classification System Using Effective Artificial Intelligence Techniques”.

This dissertation is presented by PhD student Zahraa Jafat Aliwi , under the supervision of Professor Dr. Ebtisam Abdullah and Asst. Prof. Dr. Salam Al-augby. This dissertation introduces a classification system using effective artificial intelligence techniques. This work evaluated the creation of artificial intelligence prediction models that are faster, more accurate, and more effective in diagnosis. This work helps doctors in heart diseases early detection.
Three sets of arrhythmia detection models have been built from ECG signals using different methods; To select features and take samples, the first and second models were used with the artificial minority oversampling technique to solve the unbalanced class problem (SMOTE). Based on the Fourier transform (FFT) and the wavelet transform (WT), frequency domain features were extracted, and different methods were used for classification based on automatic and deep learning. The third model, which was coded as the OWSK model, used a sequential technique that combines the One Side Selection (OSS) method to address the imbalance problem by removing overlapping, redundant samples, and noise, and WT for feature extraction, and SVM and KNN for the algorithms. Accordingly, the OWSK model achieved 90% and 98% accuracy using the intra-patient and inter-patient schemes, respectively.
As a result of this study, a web application interface was produced, which contributes to the practical application of the binary classification model. To detect cardiac enlargement based on a third dataset of chest X-ray images. The evaluation strategy through databases based on a local Iraqi database of practical clinical samples used in the testing phase of the heart hypertrophy detection model and the proposed OWSK model showed excellent and realistic results, indicating the superiority of these models in medical diagnosis compared to other models in terms of high generalization performance. Finally, this dissertation has been approved and accepted successfully.