Secure radiology image browsing tool improvised using Denoising Autoencoder with Convolutional Neural Network (DAECNN)

Computers are now considered as the daily necessities for both mankind and medical science. A doctor examines a patient, with the physical interaction and then with all the reports like scans, X-rays, blood reports, and so on. In case of Radiologist, they can’t frequently touch the screen or buttons while browsing the radiology report images, this may lead to radioactive contamination. A gesture-based browsing method is developed to overcome this issue by making the radiologist to browse the images without any close interactions with the device. An interface is provided for the surgeon where their hand-gestures are used for safe browsing of radiology report images using recent hand-gesture recognition methodologies. Further the accuracy of the system is increased by the proposed modified Convolutional Neural Network technique which uses De-noising Auto Encoder based CNN (DAECNN) to identify the hand-gesture made by the radiologist. A detailed study is made on the recent hand-gesture recognition methodologies used on secure browsing of radiology images based on accuracy. The proposed technique is compared with the existing deep learning methodologies such as CNN, Adaline (Adaptive Linear Neuron), DAE (Denoising Autoencoder) and the performances are examined. The findings of the research show that the DAECNN methodology outperforms the currently used classification techniques.