الفهرس | Only 14 pages are availabe for public view |
Abstract Modern medicine produces a large amount of data which is kept in medical databases. The extraction of relevant information and provision of scientific decision making for illness diagnosis and treatment from the database is becoming more essential. Medical data mining methods and applications based on computation intelligence, such as artificial neural networks, rough sets, fuzzy systems, and convolution neural networks, help in designing or developing automated intelligent systems which help health professionals in diagnosing health problems. The principal objective of this dissertation is to develop data mining techniques and artificial intelligence on medical data. Toward this objective, this dissertation develops three intelligent automated systems that may assist personnel related to healthcare service to diagnose new patients accurately with improved diagnostic speed., First One: Aim to early prediction system of breast cancer based on medical images (histopathology images) to diagnosis breast tumor, using mixing pre-trained deep Convolutional Neural Networks (CNN) as feature extractor, and multilevel hand-crafted features and feature selection method (PCA). Experimental results show the accuracy of the proposed method of 96.97%. The Second: Hybrid Prediction System (RS_QA) based on a rough set and quasioptimal (AQ) rule induction algorithm to predict breast cancer. The suggested experiment is performed using the Coimbra Breast Cancer Dataset (BCCD) based on sets of measures that can be obtained in routine blood tests. Using classification precision, sensitivity, specificity, and receiver operating characteristics (ROC) curves, the efficiency of our suggested approach was assessed. Experimental results indicate the highest classification accuracy (91.7 percent), sensitivity (83.3 percent), and precision (94.3) obtained by the proposed (RS_QA) model. The Third system: an expert system for calculating the Risk Factor for mortality one year after thoracic lung cancer surgery. an interesting hybrid model combining near sets with soft sets, namely soft near sets. as a system for not only predicting patient lung survival or not but also, to determine the degree of risk. The correct survival classification is done with 90.0 % accuracy. Identifying the possibility of lung cancer surgery will help the doctor and patients make a more informed decision about locating the treatment methods. |