مقالات پذیرفته شده در هفتمین کنگره بین المللی زیست پزشکی
The Application of Artificial Intelligence in Early Detection of Esophageal Cancer: A Comprehensive literature review
The Application of Artificial Intelligence in Early Detection of Esophageal Cancer: A Comprehensive literature review
Mahdi Zarei,1,*
1. Research Center for Evidence-Based Medicine, Tabriz University of Medical Sciences, Tabriz, Iran
Introduction: Esophageal cancer is the eighth most common cancer in the world and the sixth leading cause of cancer mortality. This malignancy mainly consists of two types: adenocarcinoma and squamous cell carcinoma. Most esophageal cancer patients in Western countries have adenocarcinoma, unfortunately with a five-year survival rate of less than 20%. Despite the increasing incidence of adenocarcinoma, squamous cell carcinoma is the most common type of esophageal cancer in the world, with a three-year survival rate of only about 20%. This is mainly due to late diagnosis of the disease, as more than 40% of patients are diagnosed only when metastasis has occurred. Endoscopy and biopsy with pathological examination are the basis for diagnosing esophageal cancer, but this modality is invasive, time-consuming, expensive, and highly dependent on the individual's accuracy. Therefore, our need to identify efficient diagnostic modalities has led to widespread use of artificial intelligence (AI) in early detection of this disease. Therefore, the aim of this study is to evaluate the simultaneous use of AI for endoscopic diagnosis, pathological diagnosis, and identification of relevant genes associated with esophageal cancer.
Methods: In this review study, a comprehensive search was first conducted in the PubMed, Web of Science, Scopus and Google Scholar databases. After removing duplicate and non-English articles, article abstracts were reviewed to determine their relevance to the evaluated subject. After removing irrelevant articles, the full text of the articles was examined. Their data was extracted and combined and compared with each other to achieve the final result.
Results: Machine Learning (ML) and Deep Learning (DL) algorithms have been highly regarded for creating efficient diagnosis models. The application of artificial intelligence in early detection of esophageal cancer is well established in three main areas: endoscopic-based diagnosis, pathologic-based diagnosis, and identification of related genes. The application of Supervised ML algorithms along with Narrow Band Imaging (NBI) and High-Definition White-Light Endoscopy (HD-WLE) can improve the diagnostic accuracy of existing modalities in detecting Barrett esophagus. The use of these modalities along with DL algorithms can not only increase diagnostic accuracy but also determine the location of lesions for biopsy. application of HD-WLE images with DL algorithms can also increase the diagnostic accuracy of esophageal cancer, especially in young endoscopists and mid-level practitioners. How ever, different results have been reported regarding the application of AI in Endoscopic Optical Coherence Tomography and classification of Intrapapillary Capillary Loops.
The use of artificial intelligence models for pathological classification of esophageal cancer in terms of No Dysplasia, Low Grade Dysplasia, and High Grade Dysplasia has increased the accuracy of this gold standard diagnosis method. and eventually DL models and Conventional Neural Network (CNN) can be useful in early detection of cancer by identifying microRNAs, long non-coding RNAs, and cancer related protein markers.
Conclusion: Nowadays, the use of artificial intelligence models in the detection of endoscopic, pathological, and gene-based esophageal cancer is expanding. Many studies accompany the application of these algorithms with daily modalities to increase the diagnostic accuracy of the disease in the early stages. Conducting further studies with appropriate input and training data volume and effective validity methods can identify more dimensions of this issue and pave the way for the wider application of these algorithms in early detection of esophageal cancer.