Cardiovascular disease (CVD) has caused a huge burden of disease worldwide, and accurate diagnosis and assessment of CVD has a clear significance for improving the prognosis of patients. The development of artificial intelligence (AI) and its rapid application in the medical field have enabled new approaches for the analysis and fitting of various CVD data. At present, in addition to structured medical records, the CVD field also includes a large number of non-linear data brought by imaging and electrophysiological examinations. How to use AI to process such multi-source data has been explored by a large number of studies. Therefore, this review discusses the existing ways of processing various multi-source heterogeneous data with existing artificial intelligence technologies by summarizing various existing studies, and analyzes their possible advantages and disadvantages, in order to provide a basis for the future application of AI in CVD.
ObjectiveTo construct a prognostic model of esophageal squamous cell carcinoma (ESCC) based on immune checkpoint-related genes and explore the potential relationship between these genes and the tumor microenvironment (TME). Methods The transcriptome sequencing data and clinical information of immune checkpoint genes of samples from GSE53625 in GEO database were collected. The difference of gene expression between ESCC and normal paracancerous tissues was evaluated, and the drug sensitivity of differentially expressed genes in ESCC was analyzed. We then constructed a risk model based on survival-related genes and explored the prognostic characteristics, enriched pathway, immune checkpoints, immune score, immune cell infiltration, and potentially sensitive drugs of different risk groups. ResultsA total of 358 samples from 179 patients were enrolled, including 179 ESCC samples and 179 corresponding paracancerous tissues. There were 33 males and 146 females, including 80 patients≤60 years and 99 patients>60 years. 39 immune checkpoint genes were differentially expressed in ESCC, including 14 low expression genes and 25 high expression genes. Drug sensitivity analysis of 8 highly expressed genes (TNFRSF8, CTLA4, TNFRSF4, CD276, TNFSF4, IDO1, CD80, TNFRSF18) showed that many compounds were sensitive to these immunotherapy targets. A risk model based on three prognostic genes (NRP1, ICOSLG, HHLA2) was constructed by the least absolute shrinkage and selection operator analysis. It was found that the overall survival time of the high-risk group was significantly lower than that of the low-risk group (P<0.001). Similar results were obtained in different ESCC subtypes. The risk score based on the immune checkpoint gene was identified as an independent prognostic factor for ESCC. Different risk groups had unique enriched pathways, immune cell infiltration, TME, and sensitive drugs. Conclusion A prognostic model based on immune checkpoint gene is established, which can accurately stratify ESCC and provide potential sensitive drugs for ESCC with different risks, thus providing a possibility for personalized treatment of ESCC.