MULTI DISEASE DETECTION USING DEEP NEURAL NETWORK

Abstract

This study introduces a comprehensive intelligent medical prediction framework, which combines Structured clinical data and deep convo disease-specific machine learning models The aim of this study is to create lutional neural networks to detect pneumonia in chest X-rays. The framework keeps data diversity and strengthens the predictability in multiple medical fields including chronic kidney disease, heart disease, diabetes, and so on. Disease, liver disorders, Parkinson's, diabetes and pneumonia. The experimental assessment showed a good predictive ability for most of the disease type’s categories. With the classification of chronic kidney disease, there was a near-perfect discrimination of which the area under the curve was 0.96. Good ROC-AUC values for all classifiers (1.00), indicating high separability of clinical biomarkers. Good results also obtained with ensemble and neural network techniques in the detection of Parkinson's disease. Nonlinear models showed the successfulness which gave a ROC-AUC of more than 0.97. Acoustic feature learning. Ensemble fusion was shown to aid in prediction of heart disease, and to have superior classification/stability and calibration performance compared to tree-based methods while retaining these benefits models. The neural mod works very well on biochemical interactions which are complex, namely detecting liver disorderelling and ensemble aggregation. The prediction of diabetes was still limitedly complicated due to the overlapping characteristics feature in the metabolism and that led to moderate performance in the classification, indicating the need for more complex feature representations. Medical image analysis was achieved using a deep learning model based on ResNet50V2. Chest x-ray diagnosis of pneumonia. The model shows high recall, f1-score and correctness (above 95%) and VGG19 is performing well, although not as well. Minimal overfitting, and convergence. A confusion matrix analysis also was performed and showed excellent.