A Robustness-Aware Hybrid Deep Learning Framework With Blockchain-Based Auditability For Lung Cancer Detection

Abstract

Accurate and reliable detection of lung cancer from Computed Tomography (CT) scans is essential for early diagnosis and effective clinical decision-making. Although hybrid deep learning models have demonstrated promising performance in medical image analysis, their robustness under realistic imaging degradations remains insufficiently explored. Moreover, safe storage and traceable reporting of diagnostic results are vital in, credible clinical implementation. The proposed paper is a robustness-conscious blockchain-based hybrid deep learning framework of secure lung cancer detection. The suggested system combines the HCNN-ALSTM and AHyNet systems in one experimental pipeline and explicitly simulates real-world CT degradations, such as Gaussian noise, motion blur, and partial occlusion. Relative accuracy degradation is used to develop a Robustness Index (RI) to quantitatively assess diagnostic stability. CT images are encrypted and stored off-chain, and diagnostic outputs and robustness measures are stored immutably on a block chain to provide integrity and auditing. Experimental performance on the publicly accessible LIDC-IDRI dataset , with an 80:20 traintest split with fivefold cross-validation, shows that AHyNet can perform at 98.08% accuracy on clean data, and that its performance is better under perturbed conditions, with an overall robustness index of 0.96. AHyNet, in comparison to the traditional CNN models, has a 6.7% higher classification accuracy and a 7-12% higher robustness in cases of noise, blur, and occlusions. Integration of blockchain adds little inference overhead, but maintains the possibility of real-time diagnostics. The results confirm that the proposed framework improves diagnostic accuracy and also significantly enhances robustness and security, thereby advancing reliable and clinically practical artificial intelligence systems for lung cancer diagnosis.