Publisher

ITSITS

Journal IPTEK The Journal for Technology and ScienceIPTEK The Journal for Technology and Science Abstract

Cancer is the worlds second-leading cause of death, arising from abnormal cell growth that invades the bodys cells and tissues. Simultaneous occurrences of lung and colon cancer are not uncommon, with lung cancer often emerging as the second primary cancer in colon cancer patients. While Deep Learning (DL) approaches have shown promise in accurate cancer classification, recent studies highlight the susceptibility of DL models to perturbations in input images. Merely achieving accuracy is insufficient; models must demonstrate resilience against even the slightest perturbations by applying adversarial defence methods. This study aims to enhance the reliability of the Convolutional Neural Network (CNN) algorithm in the face of adversarial attacks by implementing adversarial training. Leveraging the LC25000 dataset and various pre-trained CNN models for classification, we employ adversarial attack methods such as Carlini and Wagner, DeepFool, and SaliencyMap alongside adversarial training for defence. Evaluation metrics include precision, recall, F1-score, accuracy. Our assessment involves scrutinizing adversarial attacks and defences on histopathology images related to lung and colon issues, representing a state-of-the-art endeavour. The results indicate a significant improvement in susceptibility to adversarial attacks on histopathological images of the lungs and colon, from 0% to 81%.

Conclusion

This research successfully implemented robustness for trained models to classify lung and colon cancer histopathology data using adversarial training.The models were initially vulnerable to adversarial attacks, with accuracy dropping to 0%, but were improved through adversarial training.The best results were achieved with the GoogLeNet model, increasing accuracy on perturbed data from 0% to 81%.These findings demonstrate the effectiveness of adversarial training in enhancing the resilience of CNN models against adversarial attacks in histopathological image classification.

Future Research

Further research should investigate the application of alternative defense mechanisms, such as Defensive Distillation, Interval Bound Propagation, and Defense GAN, to further enhance model robustness. Expanding the study to include diverse datasets of histopathological images from various sources and patient populations is crucial to assess the generalizability of the proposed adversarial training approach. Future work could explore the development of novel adversarial attack strategies specifically tailored to the characteristics of histopathological images, potentially uncovering new vulnerabilities in existing CNN models. Investigating the interpretability of CNN models after adversarial training could provide insights into how the models learn to defend against attacks and identify potential biases or artifacts. Finally, exploring the integration of adversarial training with other techniques, such as data augmentation and transfer learning, may lead to even more robust and accurate cancer classification systems, ultimately improving diagnostic capabilities and patient outcomes.

References

1. Lung Cancer Classification in Histopathology Images Using Multiresolution Efficient Nets - Anjum - 2023... onlinelibrary.wiley.com/doi/10.1155/2023/7282944Lung Cancer Classification in Histopathology Images Using Multiresolution Efficient Nets Anjum 2023 onlinelibrary wiley doi 10 1155 2023 7282944

Metadata

File size	602.77 KB
Pages	12
DMCA	Report