Abstract

Deep learning is an artificial intelligence approach based on the training of convolutional neural networks (CNN) and the skull area detection factor, as it is the most powerful option for image analysis. We analyzed and evaluated the diagnostic performance of a deep learning approach to identify skull fractures and the damaged area of the brain caused by head injury in CT-scan images and also explored its clinical applicability. In this research, 3000 CT-scan images were used to determine skull fracture patients in different cases. The CNN-LSTM method with SIFT and the k-NN algorithm was used as a deep learning model. Accuracy, specificity, precision, recall, and F1-score were measured to evaluate the deep learning algorithm’s diagnostic performance. The segmentation method is used in the proposed approaches with SIFT and the k-NN algorithm method to improve the accuracy of depressed skull fracture CT-scan image datasets for the detection of damaged areas of the skull. The CNN-LSTM method is used to classify the damaged area of the skull fracture in the brain that is improved. The accuracy is 99.99%, the specificity is 99.99%, the precision is 99%, the recall is 99.1%, and the F1?score is 98.9% with less computational time (10.1 seconds) in our novel model achievement. Our proposed model successfully prepares the image for the next segmentation and classification method. The results of the experiment show that the proposed strategy is effective. The proposed model can be used to solve the depressed skull fracture detection and classification technique as a proposed hybrid model. The outcomes of the improved provides N-LSTM model provide better results. These results generate better accuracy than the previous results.