@article{Article_83,

title = {Deep Learning Techniques for Liver Cancer: A Survey on Early Prediction, Detection, and Prognosis of Metastasis and Survival},

journal = {Communications in Combinatorics, Cryptography & Computer Science},

volume = {2024},

issue = {1},

issn = { 2783-5456 },

year = {2023},

url = {http://cccs.sgh.ac.ir/Articles/2024/issue 1/1-11-Deep LearningTechniquesforLiverCancer.pdf},

author = {Arefe Pourmajidian and Javad Vahidi},

keywords = {Artificial Intelligence, Deep Learning, Early Detection, Liver Cancer, Prediction, Prognosis.},

abstract = {Background: The liver, an essential organ for digestion and removing toxins, is vulnerable to conditions such as liver cancer. Frequently, these ailments aren't identified until they've progressed significantly due to their understated initial symptoms. Although contemporary imaging techniques occasionally overlook these preliminary indications, the advent of advanced tech solutions is increasingly recognized as pivotal. The urgency of early detection is underscored by the severe repercussions of liver cancer. While traditional diagnostic measures have their limitations, nascent technologies, notably Deep Learning (DL) and especially Convolutional Neural Networks (CNN), are exhibiting remarkable diagnostic prowess, occasionally outpacing human expertise. Leveraging these DL methodologies and state-of-the-art tools has the potential to revolutionize early liver cancer detection, thus mitigating patient mortality rates and subsequently trimming both the costs and duration of treatments. Methodology: This survey assessed 84 notable articles from top journals over the last five years. Selection prioritized articles showcasing precise computations with evaluable outcomes. Our focus was on the potential of DL in early detection of precancerous liver lesions, liver cancer diagnosis, and survival rate predictions. Results: DL proved highly effective in early lesion detection and liver cancer diagnosis, achieving near-perfect accuracy of 100% in specific datasets. Prognoses concerning metastasis and survival reached an accuracy of 89.1%. Conclusion: Our survey highlights the advantage of merging neural networks with other techniques for better image classification. However, data-related constraints limit its wider application, especially due to the lack of a global standard for AI in biomedical imaging. Collaborative efforts are needed to curate extensive datasets, and more research is essential on precancerous lesions and liver cancer prognosis.}

};