Akademik Veri Yönetim Sistemi
5th International Conference on Computer Simulations in Biomedicine, Ljubljana, Slovenya, 01 Nisan 2003, cilt.7, ss.453-457
One of the current practices of diagnosing valve stenosis is based on the investigation of the Doppler Ultrasound's Fast Fourier Transformation (FFT) sonogram. This method depends on the physicians' interpretation and experience and sometimes results in false diagnosis. In this study, we have facilitated Artificial Neural Networks (ANN) that will not only simplify the diagnosis but enable the physician to make a quicker judgment about the existence of stenosis, without any hesitation. The FFT sonogram of Doppler Heart Ultrasound of a healthy person very resembles to the "M" letter. The ratios of the three points (First Systolic Peak, Endpoint of Diastole, Second Systolic Peak) in the M like curve are used as inputs to our ANN system, which is trained using Levenberg-Marquardt Method. Our system is tested on 20 patients, who were subjected for recording of the Doppler Ultrasound at mitral valves. Sixteen patients' data is allocated for training purposes and four patients were tried to see whether the ANN trained complies with physicians direct diagnosis from the FFT sonogram. The testing results was found to be compliant with physicians' findings regarding the existence and level of stenosis. The fuzzy appearance of the sonogram sometimes makes physicians suspicious about the existence of stenosis. Our technique gets around this problem by using ANN to decide and assists the physician to make the final judgment in confidence. The particular hallmark of this system is being an Add-on solution which can be coupled to current already installed Doppler ultrasounds.