Detection of trisomy 21 (down syndrome) from image using deep learning convolutional neural network method

Down syndrome is a chromosomal disorder which affects physical appearance of the person. Diagnosis of Down Syndrome is performed using expensive but accurate genetic methods. For this reason, experts diagnose the syndrome using physical appearance change of the phenotype, which is simpler and cheaper. The objective of this study is to design and implement an artificial intelligence based classification and decision support system, in order to identify people with down syndrome from the photographs of phenotypes. In the study, the classification of photos with the downies and normal persons was done by using the Matlab-Alexnet software and the Deep Learning Method (CNN). Photo dataset was obtained from the internet by open source licensing. Performance tests for the system is also performed by adding noisy photos. According to the results obtained, it was concluded that the classifier system executed tasks with a success rate of 85-90% ,in the training phase and also in performance analysis processes.