AI Platform for Confidential Cancer Identification

Inadequate early detection is considered by the medical profession as a major dilemma in cancer prevention. In addition, privacy concerns arise as more and more health care data needs to be exchanged. In this paper, we propose a pioneering privacy-protecting deep learning-assisted noninvasive cancer diagnosis approach. The Internet is used primarily for diagnostic purposes where medical information first needed to be collected via wireless channels. Securing the personal healthcare data is very important as you need to protect it from unauthorized users who are just waiting to get their hands on your private and confidential information. Gathered data gets encrypted before sending it through the channel to prevent from being theft of such useful information. Performance of proposed encryption technique is evaluated by using the correlation, entropy, comparison structure material with other security metrics including energy. We proposed using myriad methods including, model-based convolutional neural networks with magnetic resonance imaging saved over pulmonary datasets in K-fold analysis, transfer learning and fine-tuning as a part our approach to cancer detection for this study. In order to verify whether these suggested DL methods outperforms well-known machine learning techniques ( Support Vector Machine, Decision Tree, Naive Bayes and Random Forest), comprehensive tests are performed. These results suggest that the CNN-based model outperforms traditional ML algorithms and obtains an accuracy of 98.9%.