With the widespread adoption of lung cancer screening and growing public awareness, the detection rate of pulmonary nodules has increased substantially, posing new challenges for clinical management. Artificial intelligence (AI) has emerged as a powerful tool across the entire management spectrum of pulmonary nodules. Beyond improving detection sensitivity and consistency in chest radiographs and low-dose CT, AI has demonstrated promising applications in malignancy risk assessment, molecular subtype prediction, preoperative 3D planning, intraoperative navigation, and postoperative monitoring. This review summarizes recent advances in the application of AI to pulmonary nodule screening, longitudinal evaluation, pathology prediction, multi-omics integration, and perioperative management. It also discusses the technical characteristics, clinical performance, current limitations, and future prospects of various AI models. The continuous development of AI is reshaping the clinical pathway of pulmonary nodules toward more efficient and individualized care.
Cancer presents a significant global public health challenge, impacting human health on a broad scale. In recent years, the rapid advancement of big data-based bioinformatics has unveiled crucial potential in precision oncology through various omics research methods. The advent of radiomics has notably expanded the application scope of medical imaging in the field. However, due to the multi-level and multifactorial nature of tumor initiation and progression, a single omics information remains insufficient to meet the demands for advancing precision oncology strategies. Multi-omics research has become an emerging trend. The research paradigm integrating radiomics with other omics offers a novel perspective for personalized decision-making in oncology. Nevertheless, there persists a need to introduce more integrated new technologies and theories to expedite the progress of this field.
Lung cancer is the malignant tumor with the highest incidence and mortality rates worldwide, and its high lethality is primarily due to its subtle early symptoms, with most cases being diagnosed at an advanced stage. Currently, the diagnosis of lung cancer mainly relies on tissue biopsy to obtain pathological evidence, but this method has limitations such as high invasiveness, restricted sampling, and the risk of complications. Therefore, developing safe, effective, and non-invasive strategies for the early screening and diagnosis of lung cancer (stages Ⅰ/Ⅱ) holds significant clinical importance. As a key component of liquid biopsy, exosomes can stably carry a variety of biological molecular information from their cells of origin. Studies have shown that microRNAs, long non-coding RNAs, circular RNAs, and specific proteins abundantly present in exosomes exhibit abnormal expression during the development of lung cancer, demonstrating high diagnostic value. Compared to traditional detection methods, exosome-based detection offers advantages such as non-invasiveness, repeatability, ease of operation, and cost-effectiveness. This article systematically reviews recent research progress on exosomes as liquid biopsy biomarkers for the early diagnosis and screening of lung cancer, focusing on their potential clinical applications, and explores the prospects of exsomes in the early intervention, precise diagnosis, and prognosis improvement for lung cancer.
ObjectiveTo understand the research advances in molecular classification systems of gastric cancer and explore their clinical application value in precision diagnosis and treatment, as well as future development directions. MethodA literature search was conducted to identify and summarize the classic classification systems, including the Singapore-Duke typing, The Cancer Genome Atlas (TCGA) typing, and the Asian Cancer Research Group (ACRG) typing, as well as novel precision typing frameworks driven by cutting-edge technologies such as single-cell sequencing, spatial transcriptomics, epigenetics, metabolomics, and multi-omics integrative analysis. ResultsGastric cancer is characterized by high heterogeneity, and traditional pathological typing is difficult to meet the requirements of precision diagnosis and treatment. Among the classic molecular classifications, the Singapore-Duke classification classifies gastric cancer into proliferative, mesenchymal, and metabolic subtypes, and which are correlated with drug sensitivity. According to the multi-omics features, the TCGA classification categorizes gastric cancer into four subtypes: Epstein-Barr virus (EBV)-positive, microsatellite instability (MSI), genomically stability (GS), and chromosomal instabilty (CIN), among which EBV-positive and MSI subtypes are associated with the best prognosis, while the GS subtype shows the worst prognosis. The ACRG classification (Asian cohort) categorizes gastric cancer into MSI, microsatellite stable (MSS)/TP53-active, MSS/TP53-inactive, and MSS/epithelial-mesenchymal transition (EMT) subtypes, among which MSI tumors show the best prognosis, and MSS/EMT tumors the worst. Regarding novel frontier classifications, tumor microenvironment ecotypes and cancer-associated fibroblast subpopulations are identified by single-cell and spatial transcriptomics. Additionally, immune consensus subtypes, a PANoptosis-related long non-coding RNA prognostic model, and tumor immune microenvironment classifications, and other subtypes have been constructed through epigenetics, metabolomics, and multi-omics integrative analysis. In clinical translation, different molecular subtypes are matched with corresponding therapeutic strategies, and the combination of molecular classification and TNM staging is enabled to improve the accuracy of prognostic evaluation. ConclusionsMolecular classification of gastric cancer provides a stratification basis for precise diagnosis and treatment, yet its clinical translation still faces challenges such as high technical cost and intratumoral heterogeneity. In the future, relying on artificial intelligence, liquid biopsy, and other technologies, clinically practical subtype-guided individualized therapeutic strategies can be realized.