Título: Scintilogram Image Pre-Segmentation using Neural Network
Autores: Moreno, Ramon A.; Cabral Jr., Euvaldo F.
Resumo: One of the difficulties in gated blood pool images is to identify the left ventricle and calculate the ejection fraction. In this work an automatic image segmentator has been developed using a Self Organized Feature Finder (SOFF) neural network. Fifty and nine images have been used for the training and fifty nine for testing. First the program is trained to recognize the heart’s pattern and second it searches the image looking for the pattern. Once it is found a box is drawn around it. Recognition raises 69% which can be maximized with re-training of the misclassified patterns. The next step will be the automatic edge detection of the selected area and the determination of the ejection fraction.
Código DOI: 10.21528/CBRN2001-130
Artigo em pdf: 5cbrn_130.pdf
Arquivo BibTex: 5cbrn_130.bib