Prostat kanser riskinin tespitinde sınıflandırıcı tabanlı uzman sistem tasarımı
Küçük Resim Yok
Tarih
2007
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Ege Üniversitesi
Erişim Hakkı
info:eu-repo/semantics/openAccess
Özet
Prostat kanseri, erkeklerde en sık görülen ve ikinci sıklıkta ölüme neden olan bir hastalıktır. Prostat kanserine erken dönemde tanı konuldugu takdirde, ameliyat yapılabilir ve hastalık tedavi edilebilir. Bu çalısmada amaç; organa sınırlı evrede tanı konularak cerrahi tedavi sansı olan prostat kanserini olabildigince erken yakalamak için bir uzman sistem tasarlamaktır. PSA (Prostat Specific Antijen) , Serbest PSA, prostat hacmi, prostat yogunlugu, kilo, boy, BMI (beden kütle indeksi), sigara, tansiyon ve nabız gibi risk faktörlerini kullanarak en dogru tanıyı koymaktır. Bir baska deyisle, en az sayıda hastaya biyopsi yaparak, en çok sayıda kanser tanısı koymayı hedeflenmektedir. Diger amaç, sigara ve BMI faktörüyle, prostat kanseri arasındaki iliski olup olmadıgının incelemektir Tasarımlanan sistemin esas görevi kanser evresini sınıflayabilmektir. Sınıflandırma için yapay sinir aglarına ait SCG, BFGS, LM yöntemleri ile Destek Vektör Makinelerine ait Dogrusal, Radyal tabanlı fonksiyon, polinom kernel sınıflandırıcı tipleri kullanılmıstır. En iyi sonucu veren polinom kernel algoritmasının basarısı %83 tür. Ayrıca sigaranın prostat kanseri riskini artırıcı bir faktör oldugu ve BMI'nin ise olmadıgı tespit edilmistir. Anahtar Sözcükler: Uzman sistemler, Prostat kanseri, Yapay Sinir Agları, Destek Vektör Makineleri, risk faktörleri, sınıflandırıcılar
Prostate cancer is the most frequently seen disease for men and it is also the second deadly disease of men. Provided that prostate cancer can be diagnosed early, medical surgery operation can be performed so the disease can be treated. In this study, the aim is to design a classifier based expert system for early diagnosis of the organ in constraint phase. The other purpose is to reach informed decision making without biopsy by using following risc factors; PSA (Prostate Spesific Antigen), Free PSA, prostate volume, prostate density, weight, height, BMI (Body Mass Index), smoking and heart-rate. In other words, We want to diagnose cancer in optimum level where decrease the number of patients to whom applied biopsy The other purpose is to investigate a relationship between Body Mass Index and smoking factor and Prostate Cancer. The main task of the design system is to classify the early phase of prostate cancer. For classification process, the Scaled-Conjugate Gradient, BFGS and Levenberg?Marquardt training algorithms of Artificial Neural Networks and also linear,radial based function and polynominal kernel functions of Support Vector Machine (SVM) were used. The proposed system was designed with polynominal kernel function which had the best performance. The success of the system is %83. Moreover It was determined that smoking is a factor which increases prostate cancer risk but BMI was not . Keywords: Expert System, Prostate cancer, Artifical Neural Networks, Support Vector Machine, risk factors, classifiers.
Prostate cancer is the most frequently seen disease for men and it is also the second deadly disease of men. Provided that prostate cancer can be diagnosed early, medical surgery operation can be performed so the disease can be treated. In this study, the aim is to design a classifier based expert system for early diagnosis of the organ in constraint phase. The other purpose is to reach informed decision making without biopsy by using following risc factors; PSA (Prostate Spesific Antigen), Free PSA, prostate volume, prostate density, weight, height, BMI (Body Mass Index), smoking and heart-rate. In other words, We want to diagnose cancer in optimum level where decrease the number of patients to whom applied biopsy The other purpose is to investigate a relationship between Body Mass Index and smoking factor and Prostate Cancer. The main task of the design system is to classify the early phase of prostate cancer. For classification process, the Scaled-Conjugate Gradient, BFGS and Levenberg?Marquardt training algorithms of Artificial Neural Networks and also linear,radial based function and polynominal kernel functions of Support Vector Machine (SVM) were used. The proposed system was designed with polynominal kernel function which had the best performance. The success of the system is %83. Moreover It was determined that smoking is a factor which increases prostate cancer risk but BMI was not . Keywords: Expert System, Prostate cancer, Artifical Neural Networks, Support Vector Machine, risk factors, classifiers.
Açıklama
Anahtar Kelimeler
Biyomühendislik, Bioengineering, Elektrik ve Elektronik Mühendisliği, Electrical and Electronics Engineering, Üroloji, Urology, Neoplazmlar, Neoplasms, Prostat, Prostate, Risk faktörleri, Risk factors, Uzman sistemler, Expert systems
