NNLGBM: Medical Image Classification through Secure Collaboration in Pneumonia Detection by Blending NN and LGBM

Abstract

The amount of digital photos being transmitted over computer networks is increasing at an alarming rate. For a variety of reasons, including confidentiality, authenticity, and reliability, the safeguarding of digital data, particularly medical imaging, becomes increasingly critical. Digital watermarking has swiftly gained popularity as a cutting-edge technology for enhancing the security of digital photographs, and it is becoming increasingly popular. The incorporation of a watermark onto a medical image can serve to authenticate and ensure the integrity of the image. Pneumonia is responsible for the deaths of over 700,000 youngsters per year and affects approximately 7% of the world's population. When it comes to diagnosing this disease, chest X-rays are the primary tool. Examining chest X-rays, on the other hand, is a difficult assignment for even the most experienced radiologist. There is an urgent need to improve the precision with which diagnoses are made. Using computerized chest X-ray images, we present an appropriate method for the clinical condition that can be utilized to guide clinicians through the judgment process. A unique strategy based on a combination of Deep learning neural networks and LGBM in the most efficient manner is presented. A supervised learning strategy is applied in this case, in which the network anticipates the outcome depend on the merits of the data set that was used to train the network. Data augmentation approaches that enhance the training database in a balanced manner are used to increase the training dataset. The suggested classifier outperforms all of the individual models in terms of accuracy. Final assessment is made not just in terms of test accuracy but also in terms of the AUC score, which measures the overall effectiveness of the model. On a chest X-ray dataset, the final proposed weighted classifier model achieves a test accuracy of 99.43 percent and an AUC score of 99.76 percent, demonstrating superior performance. The developed framework can therefore be utilized to make a rapid diagnosis of pneumonia and to assist physicians in their diagnostic procedure.