Feed Forward Back Propagation Neural Network Based Computer Aided Detection and Diagnosis of Malignant Regions in Thyroid Images
Amongst numerous thyroid nodules, it is necessary to classify as several malignant nodules from benign nodules using Fine Needle Aspiration biopsies. Medical image investigation has played a vital task in numerous clinical measures for detecting special types of human diseases especially thyroid malignancy. Computer aided analysis aid radiologists to progress the diagnosis accuracy, decrease biopsy ratio and put aside their time. Feed Forward Back Propagation Neural Network classifier is planned to classify the thyroid nodules as normal or abnormal. To eliminate the artifacts within the ultrasound images Gaussian filter is used. The extracted thyroid ultrasound image features are classified as malignant and benign nodules with the help of proposed classifier. The experimental result achieves good classification performance, attaining 99.77% sensitivity, 99.23% specificity and 98.20% classification accuracy.
Keywords: Computer aided diagnosis, Features extraction, Feed Forward Back Propagation Neural Network, Malignant, Benign