Presentation is loading. Please wait.

Presentation is loading. Please wait.

The Fetal Medicine Foundation Computational Intelligent Diagnostic System in Predicting Fetal Aneuploidies Christos Schizas Kypros Nicolaides Andreas Neocleous.

Similar presentations


Presentation on theme: "The Fetal Medicine Foundation Computational Intelligent Diagnostic System in Predicting Fetal Aneuploidies Christos Schizas Kypros Nicolaides Andreas Neocleous."— Presentation transcript:

1 The Fetal Medicine Foundation Computational Intelligent Diagnostic System in Predicting Fetal Aneuploidies Christos Schizas Kypros Nicolaides Andreas Neocleous Kleanthis Neokleous Natasa Schiza Costas Neocleous FMF, University of Cyprus, Cyprus University of Technology, Cyprus Computational Intelligence Artificial neural networks Artificial neural networks Evolutionary systems / Evolutionary systems / Genetic algorithms Genetic algorithms Artificial immune systems Artificial immune systems Fuzzy systems Fuzzy systems Maternal age Previous trisomy Crown-rump length Gestational age Nuchal translucency Fetal heart rate Free ß-hCG PAPP-A Nasal bone Tricuspid flow Ductus venosus flow

2 The Fetal Medicine Foundation Computational Intelligent System in predicting fetal aneuploidies Objective: Employ computational intelligence to predict fetal aneuploidies All data: Total singleton pregnancies 34,182 Euploid33,792 (98.8%) Aneuploidy 390 (1.2%) Trisomy Trisomy Trisomy Triploidy 18 Turner syndrome 35 Data for training, simulations and validations: Training various artificial neural networks 26,000 Totally unknown cases used for validations 8,182 Artificial Neural Network Architecture Input (10 neurons) Age, previous trisomy, CRL, NT, FHR, ß-hCG, PAPP-A, NB, TR, DV (Linear activation) Hidden Layer 1 (80 neurons) (Logistic activation) Output Layer (5 neurons) Normal / Abnormal (Turner, T13,T18,T21) (Logistic activation) Hidden Layer 2 (10 neurons) (Symmetric logistic activation) Hidden Layer 3 (80 neurons) (Logistic activation)

3 The Fetal Medicine Foundation Results on the unknown validation (verification) data set: Predicted Correct ALL cases EuploidAneuploid 8, ,017 (99.8%) 64 (100%) Classification into EUPLOID - ANEUPLOID Normal Trisomy 21 Trisomy 18 Trisomy 13 TriploidyTurner ALL cases 4,5214, Predicted Correct 4,505 (99.5%)4,481(99.98%)18(85.7%)6(60.0%)000 Classification into EUPLOID - T21 – T18 - T13 – Triploidy - Turner Computational Intelligent System in predicting fetal aneuploidies Predicted Correct ALL cases Euploid Trisomy 21 8, ,016 (99.8%) 54 (90.0%) Classification into EUPLOID – Trisomy 21

4 The Fetal Medicine Foundation Conclusions  There is a very good discrimination between Euploid and Aneuploid cases  There is a good discrimination between normal and Trisomy 21 cases  T13, Triploidy and Turner cases are hard to predict (mainly because of the small number of cases available for network training) Thank you Computational Intelligent System in predicting fetal aneuploidies


Download ppt "The Fetal Medicine Foundation Computational Intelligent Diagnostic System in Predicting Fetal Aneuploidies Christos Schizas Kypros Nicolaides Andreas Neocleous."

Similar presentations


Ads by Google