No abstract

The heart anomalies detection is a significant task in cardiac medical research. The CT, ULTRASOUND, CTA and MRI scans have been used to detect heart diseases but giving false experimental outcomes in longer time of conversion (ToC). Therefore, patients haven’t getting better treatment from doctors. So that in this research work an ultrasound image scan-based heart disease prediction and classification is performed with deep learning technology. The LeNet 10 deep learning classifier has been trained Kaggle dataset using appropriate CNN layers. Proposed CNN LeNet -10 is a 165 layers technology consists of flattened layer, dense layer, convolution layer, max pooling layer and etc. Classification and feature extraction has been performed to loading with LeNet-10 architecture. The real time heart ultrasound test images are collecting from Manipal super specialty hospital Vijayawada, these test features are managed to test.CSV file. In pre-processing step, Ostu segmentation and histogram equalization is applied to make heart ultrasound images to be clear. In Segmentation, edge and region-based convolutional steps are applied such that deep features have been identified. LeNet-10 classification is used to find affected area as well as abnormality location. Finally proposed deep learning with confusion matrix can generating application measures. Implementation has been performed on python 3.9 and DL (Deep learning) packages like TensorFlow, keras, sklearn and etc. The measures like Accuracy 98.37%, sensitivity 97.81%, Recall 98.34% and F1 score 98.98% had been attained, proposed heart disease estimation application is more robust and compete with present technology.

No abstract

No abstract

No abstract