This article is mentioned in:

Introduction

According to the Global Cancer Statistics 2020, lung cancer is the second most common cancer and the leading cause of cancer-related death (1). The distinction of pulmonary nodules and early diagnosis of lung cancer are crucial for improving prognosis and enabling surgical intervention. However, current screening methods, such as imaging and tissue biopsy, have limitations, such as false positives, radiation exposure and potential complications like hemorrhage, infection, pneumothorax and implantation metastasis. Therefore, there is an urgent need for more reliable, testable and safer biomarkers. Liquid biopsy has emerged as a non-invasive examination that can reveal important tumor features, including gene mutations and metabolic changes. It is increasingly recognized as a valuable alternative for diagnosing and differentiating pulmonary nodules. The present review aimed to discuss the current status of pulmonary nodule diagnosis and summarize the latest applications of liquid biopsy in identifying malignant and benign pulmonary nodules (BPN) (Table I; Fig. 1).

Figure 1 Combined use of multi-biomarkers in pulmonary nodules and lung cancer. Liquid biopsy serves as a crucial complement to traditional diagnostic methods in the identification and treatment of malignant and benign pulmonary nodules. CT, computerized tomography; PET, positron emission tomography; MRI, magnetic resonance imaging; AI, artificial intelligence; CNNs, convolutional neural networks; EVs, extracellular vesicles; TEPs, tumor-educated platelets; ncRNA, non-coding RNA; ctDNA, circulating tumor DNA; TAAb, autoantibodies targeting tumor-associated antigens.

Table I Application of liquid biopsy in the diagnosis of pulmonary nodules.

Table I

Application of liquid biopsy in the diagnosis of pulmonary nodules.

Class	Author/s, year	Biomarkers	Method	Deregulation	Diagnosis and differential diagnosis	Targets or biological processes	Sensitivity, %	Specificity, %	AUC	Other applications and advantages	(Refs.)
EV-associated miRNAs	Gao et al 2022	miR-106b-3p+ miR-125a-5p+ miR-3615+ miR-450b-5p	Sequencing RT-qPCR	NA	LUAD vs. HC and BPNs	ZNRF3+ KRAS and NF-kB pathways+ NA+YAP1	84.9	80.9	0.902	Recognizing patients with AIS and MIA	(31)
	Zhong et al 2021	miR-520c-3p+ miR-1274b	TLDA+ RT-qPCR	Up	NSCLC vs. BPNs	AKT1, AKT2+ CNN1	NA	NA	0.823	The highest efficiency of EV extraction and low albumin impurities	(37)
	Zhang et al 2023	miR-1290/miR-29c-3p	Sequencing RT-qPCR	Up: miR-1290 Down: miR-29c-3p	LC vs. BPNs	IRF2/PDGF-BB/wnt pathway	97.37/89.47	89.47/84.21	0.934/0.868	Distinguishing between NSCLC and SCLC, and diagnosing early-stage lung cancer	(30)
	Chen et al 2022	miR-4732-5p+ miR-451a+ miR486-5p+ miR139-3p	Fucose-capture sequencing, RT-qPCR	Up: miR-4732-5p/miR-451a/miR486-5p Down: miR139-3p	LUAD vs. BPNs	XPR1+ATF2+ ribosomal proteins+ CHEK1	91.07	66.36	0.855	Developing a new method for extracting EV	(43)
	Zheng et al 2023	miR-4497	Sequencing RT-qPCR	Down	NSCLC vs. BPNs	GBX2	73.30	72.60	0.748	Diagnosing early-stage NSCLC and monitoring the degree of tumor malignancy	(48)
	Yang et al 2020	miR-21/Let-7a ratio	RT-qPCR	Up	NSCLC vs. BPNs	PTEN	56.00	82.61	0.754	NA	(36)
EV-associated long RNAs	Zhang et al 2023	SEC62+ ANXA4+ KIAA1217+ TMTC1+ KAZN+ AC009303.2+ HLA-E+ ARHGAP30+ CICP3+ CXCL8+ IVNS1ABP+ MTCYBP18+ CSRNP1+ GFI1B+ SNX29+ GOLGA3+ USP49+ RNU6-959P + SPIRE1 + LRRC49 + TMEM231 + PLTP+NPC2	Sequencing high-through-put screening and support vector machine analyses	Up: SEC62/ANXA4/KIAA1217/TMTC1/KAZN/AC009303.2 Down: Others	LUAD vs. BPNs	Integrin α/CAV1 pathway etc.	93.75	85.71	0.918	High diagnostic accuracy and classification of AIS vs. MIA/IAD	(51)
	Min et al 2022	RP5-977B1	Deep sequencing RT-qPCR	Up	NSCLC vs. HC	NA	82.86	84.93	0.890	Diagnosing early-stage NSCLC and predicting the prognosis of NSCLC	(53)
EV-associated protein	Kuang et al 2019	FGG+FGB	WB	Up	LC vs. BPNs	Adherence of tumor cells	81.40	70.00	0.794	NA	(56)
	Chang et al 2023	Versican	WB and ELISA	Up	NSCLC vs. HC and BPNs	Formation of an inflammatory tumor microenvironment	85.45	61.82	0.790	Positive correlation with TNM stage, lymph node metastasis, distant metastasis and mutation	(57)
	Yang et al 2023	IGHV4-4+ IGLV1-40	WB and ELISA	Up	NSCLC vs. HC	NA	88.73	85.00	0.930	Diagnosing metastatic NSCLC and combining traditional bio-markers to improve diagnostic efficiency	(58)
	An et al 2019	Fibronectin	LC-ESI MS/MS	Up	NSCLC vs. HC	mTOR pathway	NA	NA	0.844	Increasing the tissue specificity	(61)
TEPs	Li et al 2021	Linc-GTF2H2-1+ RP3-466P17.2+ lnc-ST8SIA4-12	qPCR	Up: Lnc-ST8SIA4-12 Down: Linc-GTF2H2-1/RP3-466P17.2	LC vs. HC	NA	82.60	87.10	0.921	Diagnosing early-stage LC and combining traditional biomarkers to improve diagnostic efficiency	(64)
	Xing et al 2019	ITGA2B	RNA-seq and PCR	Up	NSCLC vs. BPNs	Hematogenous metastasis of cancer	96.40	81.70	0.940	Predicting overall survival	(65)
miRNAs	Xi et al 2018	miR-17+ miR-146a+ miR-200b etc.	RT-qPCR	Up	NSCLC vs. BPNs	STAT3+ IRS2+ RhoE etc.	54.8-83.3	60.00-86.70	0.675-0.740	NA	(71)
	He et al 2018	miR-199a-3p+ miR-148a-3p+ miR-210-3p+miR-378d+ miR-138-5p	RT-qPCR	NA	LC vs. BPNs	Rheb etc.	34.00	90.20	NA	NA	(72)
	Shen et al 2011	miR-21+ miR-210+ miR-486-5p	RT-qPCR	Up: miR-21/miR-210 Down: miR-486-5p	LC vs. BPNs	PTEN+ JAK2/STAT3 pathway+ TGF-β/SMAD2 signaling pathway	75.00	84.95	0.860	NA	(73)
	Fan et al 2018	Five paired miRNA ratios (miR-15b-5p/miR-146b-3p, miR-20a-5p/miR-146b-3p, miR-19a-3p/miR-146b-3p, miR-92a-3p/miR-146b-3p, and miR-16-5p/miR-146b-3p)	RT-qPCR	Up	NSCLC vs. BPNs	NA	70.00	90.00	0.870	NA	(74)
lncRNAs	Chen et al 2023	lncRNA THRIL	RT-qPCR	Up	LC vs. BPNs	miR-99a	87.34	83.78	0.912	Exploring the role of THRIL in the development of LC	(77)
	Jiang et al 2018	lncRNA XLOC_ 009167	RT-qPCR	Up	LC vs. HC	NA	78.70	61.80	0.740	Remains stable in whole blood under different conditions	(78)
pfeRNAs	Liu et al 2021	pfeRNAa to pfeRNAh	RT-qPCR	NA	NSCLC vs. BPNs	NA	77.10	74.25	0.840	Classification with vs. without pulmonary nodules	(81)
ctDNA mutation	Jiang et al 2021	RNF213	NGS and IHC	Up	LC vs. BPNs	α-KGDD	NA	100.00	NA	High specificity	(84)
ctDNA methylation	Liang et al 2019	9 hypermethylated markers	High through-put DNA bisulfite sequencing	Up	LC vs. BPNs	NA	79.50	85.20	NA	Diagnosing early-stage LC	(87)
	Chen et al 2020	CDO1+ SOX17+ HOXA7	MOB and qMSP	Up	NSCLC vs. BPNs	Metabolism of cysteine+ Wnt/β-catenin signaling pathway+ NA	90.00	71.00	0.880	Using a new approach to extract DNA	(90)
	Fang et al 2022	Methylation level in FUT7 CpG_1-7	Mass spectrometry	Down	LC vs. BPNs	EMT and immune infiltration	71.43	56.65	0.658	Large number of clinical samples and classification LC vs. HC	(88)
Antibodies	Lastwika et al 2019	FCGR2A+ EPB41L3+ LINGO1+ S100A7L2	High-density protein arrays and ELISA	Up	NSCLC vs. BPNs (with the largest diameter 8-20 mm)	NA	91.70	57.10	0.780	Potentially useful for immunotherapeutic target selection	(97)
	Auger et al 2023	Annexin2+ DCD+ MID1IP1+ PNMA1+ TAF10+ ZNF696	High-through put protein microarrays and custom Luminex immunobead assays	NA	'Actionable' vs. 'non- actionable' nodules	Tumor metastasis etc.	97.20	50.00	NA	Using a new approach to assay the candidate antibodies and high sensitivity	(99)
	Shome et al 2023	20 anti- microbial antibodies	Protein microarrays and ELISA	NA	LUAD vs. BPNs	NA	NA	NA	0.800	Large number of amples profiled sagainst hundreds of bacterial and viral antigens	(100)
Combined indicators	Wang et al 2023	Plasma PGM5-AS1, SFTA1P, CTA-384D8.35, Log10CEA, Log10CA125, SCC and NSE	RT-qPCR	NA	NSCLC vs. HC	miRNA-423-5p, Hippo-YAP/TAZ signaling	NA	NA	0.970	Large number of clinical samples, outstanding diagnostic performance and clinical applicability	(52)
	Li et al 2019	lncRNA GAS5+CEA	RT-qPCR	NA	NSCLC vs. HC	IGF-1R	89.06	90.00	0.929	Diagnosing early-stage NSCLC and high specificity	(54)
	Peng et al 2019	Age of the patients, ctDNA mutations and serum biomarkers	Ultra-deep sequencing	NA	LC vs. BPNs	NA	80.00	99.00	NA	Excluding the noisy background of the cfDNA and high specificity	(85)
	Xing et al 2021	PTGER4, RASSF1A, SHOX2 and the diameter of pulmonary nodules	qMSP	NA	LC vs. BPNs	Cell cycle and apoptosis etc.	89.50	95.40	0.948	High diagnostic accuracy	(93)
	He et al 2023	100-feature cfDNA methylation, clinical features and CT imaging features	Targeted DNA methylation sequencing	NA	LUAD vs. BPNs	NA	98.00	0.500	0.900	Large number of clinical samples, high sensitivity, diagnosing early-stage LC and decision-making regarding surgery	(94)
	Xu et al 2022	7AABs (P53, PGP9.5, SOX2, GAGE7, GBU4-5, MAGEA1 and CAGE), clinical information and imaging data	ELISA	Up	LC vs. BPNs	TIF1γ etc.	96.40	79.10	0.960	Good performance and repeatability	(98)

[i] SEC62, SEC62 homolog, preprotein translocation factor; ANXA4, annexin A4; TMTC1, transmembrane and tetratricopeptide repeat containing 1; KAZN, Kazrin; HLA-E, human leukocyte antigen E; ARHGAP30, Rho GTPase-activating protein 30; CXCL8, C-X-C motif chemokine ligand 8; IVNS1ABP, influenza virus NS1A binding protein; MTCYBP18, mitochondrial cytochrome B pseudogene 18; CSRNP1, cysteine-serine-rich nuclear protein 1; GFI1B, growth factor independent 1B; SNX29, sorting nexin 29; GOLGA3, golgin A3; USP49, ubiquitin specific peptidase 49; RNU6-959P, RNA U6 small nuclear 959 pseudogene; SPIRE1, spire type actin nucleation factor 1; LRRC49, leucine-rich repeat-containing protein 49; TMEM231, transmembrane protein 231; PLTP, phospholipid transfer protein; NPC2, Niemann-Pick disease type C2; FGB, fibrinogen beta chain; FGG, fibrinogen gamma chain; IGHV4-4, immunoglobulin heavy variable 4-4; IGLV1-40, immunoglobulin lambda variable 1-40; ITGA2B, tumor-educated blood platelets integrin alpha 2b; RNF213, ring finger protein 213; CDO1, cysteine dioxygenase type 1; HOXA7, homeobox A7; SOX17, SRY-box transcription factor 17; FUT7, fucosyltransferase 7; FCGR2A, Fc gamma receptor IIa; EPB41L3, erythrocyte B membrane protein band 4.1 like 3; LINGO1, nogo receptor-interacting protein-1; S100A7L2, S100 calcium binding protein A7 like 2; DCD, dermcidin; MID1IP1, MID1 interacting protein 1; PNMA1, paraneoplastic antigen MA1; TAF10, TATA-box binding protein associated factor 10; ZNF696, zinc finger protein 696; GAS5, growth arrest-specific 5; PTGER4, prostaglandin E receptor 4; RASSF1A, ras association domain family 1A; SHOX2, short stature homeobox gene two; P53, tumor protein 53; PGP9.5, protein gene product 9.5; GAGE7, G antigen 7; GBU4-5, ATP-dependent RNA helicase 4-5; MAGEA1, melanoma-associated antigen A1; CAGE, cancer-associated gene; AIS, adenocarcinoma in situ; MIA, minimally invasive adenocarcinoma; IAD, invasive adenocarcinoma; ZNRF3, zinc and ring finger 3; YAP1, Yes1-associated transcriptional regulator; CNN1, calponin h1; IRF2, interferon regulatory factor 2; PDGF-BB, platelet-derived growth factor-BB; XPR1, xenotropic and polytropic retrovirus receptor 1; ATF2, activating transcription factor-2; CHEK1, checkpoint kinase 1; GBX2, gastrulation brain homeobox 2; PTEN, phosphatase and tensin homolog; STAT3, signal transducers and activators of transcription; IRS2, insulin receptor substrate 2; RhoE, Rho family GTPase 3; Rheb, ras homolog enriched in brain; α-KGDD, α-KG-dependent dioxygenases; IGF-1R, insulin-like growth factor 1 receptor; TIF1γ, transcriptional intermediary factor 1γ; TLDA, TaqMan low density array; NGS, next-generation sequencing; IHC, immunohistochemical; AUC, area under the curve; MOB, nanoparticle-based DNA extraction; qMSP, quantitative methylation-specific PCR; EMT, epithelial-mesenchymal transition; NA, not applicable.

Current status in the diagnosis of pulmonary nodules

Pulmonary nodules, defined as rounded or abnormally cloudy lesions smaller than 30 mm in diameter, can be well or poorly demarcated and surrounded by an inflated lung on radiological imaging (2). They are classified as benign or malignant, with significant prognostic differences (3). Thus, improving existing methods or exploring new approaches for early diagnosis of pulmonary nodules is crucial. Tissue biopsy and imaging techniques, such as computerized tomography (CT), positron emission tomography (PET)/CT and magnetic resonance imaging (MRI), are currently the main diagnostic methods. The application of artificial intelligence (AI) tools on radiology images and digital pathology (4) can effectively enhance diagnostic efficiency, optimize treatment, evaluate prognosis and ultimately reduce mortality (5).

CT assessment of pulmonary nodules

Various CT characteristics (nodule size, growth rate, location, morphology, enhancement) are associated with the nature of pulmonary nodules (6). In cases where the probability of malignancy is low, such as with constant, perifissural, well-circumscribed nodules, satellite nodules with benign imaging features (stippled, laminated, dense central or popcorn patterns of calcification), or individual nodules without any risk factors (a mixed ground glass opacity <6 mm in diameter), further imaging follow-up may not be necessary if these features remain unchanged for two years or more. However, with nodules larger than 10-20 mm, the risk of malignancy increases significantly and medical follow-up is important (7). It should be noted that CT imaging for screening pulmonary nodules has limitations, as it can detect both invasive and inert tumors (8). Due to the high sensitivity of low-dose computed tomography (LDCT), numerous non-neoplastic solitary pulmonary nodules (SPN) are also detected, resulting in an increase in false positives and leading to follow-up repeat CT scans and potential issues with radiation exposure (9).

PET/CT assessment of pulmonary nodules

According to the research conducted by Dalli et al (10), PET/CT is a highly precise method for distinguishing between benign and malignant SPN. Similarly, 18F-fluorodeoxyglucose PET/CT is more accurate than helical dynamic CT for diagnosing SPN. However, PET/CT may not be reliable for qualitative assessment and staging of pure ground-glass nodular lung adenocarcinoma (LUAD) (6).

MRI assessment of pulmonary nodules

MRI holds promise for longitudinally assessing lung disease and function, offering an alternative to LDCT in patients with lung cancer. With excellent soft tissue contrast and high spatial resolution, MRI provides both morphologic and functional information through diffusion-weighted and perfusion sequences (11). However, a critical challenge in lung MRI is its susceptibility to motion artifacts caused by cardiac and respiratory movements. To mitigate these effects, implementing breath-gated or breath-holding techniques during image acquisition is recommended (12).

AI diagnosis systems for pulmonary nodules

New techniques such as computational radiomics and deep learning-based AI show promise in differentiating between malignant and benign nodules (6). Computer-aided diagnosis systems traditionally rely on imaging characteristics, including image segmentation, feature value extraction and classification, and are now being enhanced with convolutional neural networks (CNNs)-based models. These models use a multi-view strategy to improve sensitivity for pulmonary nodules. CNN models are widely employed not only in CT images but also in cytopathological (13) and histological imaging (14). Certain studies utilize CNN models to identify key gene mutations or investigate the development mechanisms of cancer (15,16). However, nodule features may influence the predictive performance of CNN models. In addition, training complex CNN models with limited training sets can lead to overfitting (14).

Liquid biopsies for pulmonary nodules

Liquid biopsy is a non-invasive and reproducible method for real-time monitoring of tumors, allowing for the rapid retrieval of pathological information from patient body fluids. It offers valuable insights into the molecular mechanisms involved in tumorigenesis and progression, making it a promising alternative to tumor tissue samples in clinical settings (17). Various biomarkers, including extracellular vesicles (EVs), circulating tumor DNA (ctDNA), platelets, plasma RNA and circulating antibodies, have been identified as non-invasive markers for diagnosing pulmonary nodules (18). The integration of advanced technologies has further enhanced the efficient capture of biomarkers for liquid biopsy, revolutionizing clinical decision-making at various stages of lung cancer management (18-21).

EVs

EVs in lung cancer

EVs, which are endosome-derived nano vesicles (30-1,000 nm) involved in intercellular communication, play a role in the immune escape, metastasis, metabolic reprogramming and drug resistance of lung cancer (22). For instance, a study found that cancer cell-derived EVs containing circUSP7 inhibited the function of CD8+ T cells, promoting the progression of non-small cell lung cancer (NSCLC) and resistance to anti-programmed death 1 therapy (23). EVs carrying snail-1 released by cancer-associated fibroblasts induce epithelial-mesenchymal transition (EMT) (24). Cancer cells can utilize glycolysis and glutaminolysis to meet their metabolic needs (25). EVs carrying circSHKBP1 enhance glycolysis by sponging microRNA (miR)-1294, leading to increased expression of the glycolytic enzyme pyruvate kinase isozyme type M2 (PKM2) and ultimately affecting the function of NSCLC cells and macrophages (26). EVs derived from cancer-associated fibroblasts contained long intergenic non-coding RNA (LINC)01614, which enhanced glutamine uptake in lung cells by upregulating the expression of glutamine transporters (27). In the study of therapeutic resistance, hypoxia-induced EVs were found to transmit cisplatin resistance to sensitive NSCLC cells by delivering PKM2 (28). In addition, EV transfer of wild-type EGFR was also shown to promote resistance to the drug Osimertinib (29). Therefore, targeting the secretion and transfer of specific cargo in EVs, such as PKM2 and EGFR, may be a promising approach to overcome treatment resistance.

Application of EVs in the diagnosis of pulmonary nodules

i) EV-associated miRNAs: Numerous studies on EV miRNA have primarily focused on the early diagnosis of lung cancer. EV miR-29c-3p and miR-1290 have been identified as superior diagnostic biomarkers for distinguishing between lung cancer and benign BPN. Their expression levels show significant differences in lung cancer, with miR-29c-3p decreased and miR-1290 elevated. These miRNAs exhibit high sensitivity (89.47 and 97.37%), specificity (84.21 and 89.47%) and area under the curve (AUC) values (0.868 and 0.934) in discriminating lung cancer from BPNs. Furthermore, they demonstrate a strong discriminative ability between NSCLC and SCLC with AUC values of 0.842 and 0.810 (30). Utilizing multiple EV miRNAs enhances diagnostic accuracy. A diagnostic signature consisting of four EV-derived miRNAs (miR-106b-3p, miR-125a-5p, miR-3615 and miR-450b-5p) has been developed for the early detection of LUAD. In training cohorts, the signature exhibited an AUC value of 0.917, a sensitivity of 83.8% and a specificity of 87.1%. These diagnostic capabilities were further validated in test cohorts (31). Mechanistic studies have elucidated the roles of these miRNAs. EV miR-106b targets phosphatase and tensin homolog, promoting migration and invasion of lung cancer cells (32). In NSCLC, miR-125a-5p inhibits the expression of histone methyltransferase Suv39H1, leading to cancer suppression both in vitro and in vivo (33). Furthermore, miR-125a-5p exerts a tumor-inhibiting effect by targeting STAT3 (34). As the target RNA of two competing endogenous long non-coding (lnc) RNAs, miR-450b-5p demonstrates a tumor-suppressive function in NSCLC (35). Not only the quantity, but also the ratio of EV miRNAs, plays a significant role in distinguishing between benign and malignant pulmonary nodules. For instance, the miR-21/Let-7a ratio is markedly elevated in NSCLC compared to BPNs, with an AUC of 0.754 and a specificity of 82.61% (36).

Zhong et al (37) utilized TaqMan Low Density Array and reverse transcription-quantitative qPCR to identify and validate a distinct pattern of circulating EV miRNAs at multiple medical centers. They found that miR-520c-3p and miR-1274b were significantly elevated in patients with NSCLC compared to healthy controls and those with benign nodules. The developed panel effectively differentiated NSCLC from benign nodules, with an AUC of 0.823, and was identified as an independent risk factor for NSCLC (37). Various intriguing findings warrant further exploration. miR-520c-3p is a tumor suppressor miRNA in NSCLC, regulating biological processes such as IL-8 (38) and PI3K/AKT signaling (39). It is higher in non-tumor tissues than in LUAD tissues. On the other hand, miR-1274b, associated with tumor growth and development (40), is upregulated in LUAD side population cells. However, as an apparent contradiction, both miR-520c-3p and miR-1274b were found to be upregulated in circulating EVs of patients with NSCLC compared to healthy controls and those with benign nodules. The upregulation of tumor-promoting miRNAs in both tissues and blood EVs is well understood, while it remains elusive how tumor-suppressive miRNAs are decreased in tumor cells but elevated in EVs. Endogenous RNAs, including miRNA targets, play a role in sorting miRNAs to EVs. The complex interactions between miRNAs and their targets can lead to the upregulation of both tumor-suppressive and tumor-promoting miRNAs in EVs (41). In addition, tumor-suppressive miRNAs in tumor cells could be eliminated through EVs and target tumor-associated immune cells in the microenvironment to promote immunosuppressive effects and induce tumor-promoting phenotypes. This dual mechanism also explains the elevation of tumor-suppressive miRNAs in EVs (42).

The main challenge in using EV-associated miRNAs for diagnosing lung nodules is the difficulty in effectively capturing and enriching tumor-derived EVs from complex blood systems. A recent study introduced a Glyexo-capture strategy using lentil lectin-magnetic beads to target exosome membranes. This dual-target method enhances the detection of valuable exosomal biomarkers with increased sensitivity and specificity. For instance, a miRNA panel consisting of miR-4732-5p, miR-451a, miR-486-5p and miR-139-3p showed promise in screening for early LUAD from BPNs, achieving an AUC of 0.855 with 91.07% sensitivity and 66.36% specificity (43). These miRNAs play crucial roles in inhibiting lung cancer through various pathways: miR-4732-5p targets xenotropic and polytropic retrovirus receptor 1 to suppress LUAD migration and metastasis (44), miR-451a inhibits invasion by targeting activating transcription factor-2 (45), miR-486-5p hinders the mTOR pathway by targeting ribosomal proteins (46) and miR-139-3p reduces lung squamous carcinoma viability by targeting checkpoint kinase 1 (47). In addition, exosomal miR-4497 is also a tumor suppressor marker, showing diagnostic efficacy in distinguishing NSCLC from BPNs with 73.3% sensitivity, 72.6% specificity and an AUC of 0.748. Importantly, miR-4497 demonstrates potential for monitoring tumor malignancy [size, tumor node metastasis (TNM) stage and metastasis] and overall survival (48).

ii) EV-associated long RNAs (exLRs): While previous studies have primarily focused on miRNAs, their limited presence in EVs hinders their effectiveness as biomarkers for lung cancer diagnosis (49). By contrast, exLRs, such as mRNAs, circRNAs and lncRNAs, are abundant in peripheral blood EVs and have shown promise as diagnostic biomarkers of lung cancer (50). Specifically, a panel of 23 exLRs, including 6 upregulated and 17 downregulated genes, identified in EVs can differentiate LUAD from BPNs with high sensitivity (93.75%), specificity (85.71%), and accuracy (88.24%). Furthermore, a signature of 17 exLRs, comprising 2 upregulated and 15 downregulated genes, can effectively distinguish between adenocarcinoma in situ and minimally invasive or invasive adenocarcinoma with exceptional sensitivity (93.33%), specificity (98.00%) and accuracy (96.25%) (51). In a recent study, a diagnostic model incorporating three exLRs (PGM5-AS1, SFTA1P and CTA-384D8.35) showed a strong predictive ability for NSCLC, with an AUC of 0.97 (52). Additionally, researchers found that EV lncRNA RP5-977B1 expression was elevated in NSCLC compared to healthy controls and patients with pulmonary tuberculosis, showing superior discriminatory power (AUC 0.890) over traditional markers carcino-embryonic antigen (CEA) (0.761) and cytokeratin 19 fragment (CYFRA21-1) (0.670). This comparative advantage was also observed in distinguishing early-stage NSCLC from controls (53). To improve diagnostic accuracy, a novel index combining gasdermin 5 and CEA was developed, achieving an AUC of 0.929, sensitivity of 89.06% and specificity of 90.00% for NSCLC diagnosis (54).

iii) EV-associated protein: EV-associated proteins, such as fibrinogen beta chain (FGB) and fibrinogen gamma chain (FGG), have been implicated in EMT progression of lung cancer (55). Researchers have demonstrated the utility of FGB and FGG levels in plasma EVs as diagnostic biomarkers for distinguishing benign and malignant pulmonary nodules. Compared to benign nodules, patients with lung cancer exhibited elevated FGB and FGG expression levels. When used individually, FGB showed a sensitivity of 0.628, specificity of 0.800 and AUC of 0.741, while FGG had a sensitivity of 0.535, specificity of 0.850 and AUC of 0.659. Combining FGB and FGG detection improved the diagnostic accuracy, with a sensitivity of 0.700 and AUC of 0.794, suggesting that these proteins could serve as sensitive biomarkers for distinguishing benign from malignant pulmonary nodules (56).

Plasma EV versican, a chondroitin sulfate glycoprotein was found to be significantly elevated in patients with NSCLC, with expression levels correlating with TNM stages and clinical parameters. Combining plasma versican and plasma EV versican showed superior diagnostic performance in identifying patients with NSCLC and those with metastasis compared to traditional biomarkers [neuron-specific enolase (NSE), CYFRA21-1 and squamous cell carcinoma antigen] (57). Yang et al (58) identified differential expression of immunoglobulin heavy variable 4-4 and immunoglobulin lambda variable 1-40 in serum EVs of patients with NSCLC, with the combination showing a high diagnostic capacity with a sensitivity of 88.73%, a specificity of 85.00% and AUC of 0.93. Recent research identified 150 altered EV proteins in patients with NSCLC, primarily involved in cell adhesion, differentiation, motility and osmoregulation, suggesting their potential as biomarkers for early NSCLC diagnosis [panel of FGB, FGG and von Willebrand factor] and metastasis prediction (panel of complement factor H related protein 5, complement component 9 and mannose-binding lectin 2) (59).

While EV biomarkers offer numerous advantages, challenges persist in their clinical application. EVs can be secreted by various cells and those from cells with pathological changes may be overshadowed by those from normal cells. Multi-level screening is an important strategy for enhancing tumor relevance and tissue specificity of EV markers. In a study analyzing EV proteomes from paired tumor and adjacent tissues, exclusive proteins HIV-1 Tat interactive protein 2 and methyltransferase like 1 were identified as specific markers for LUAD. Furthermore, comparing plasma and tissue-derived EV proteins revealed plasma EV signatures that could distinguish between lung cancers even at early stages (60). In another study using lung cancer serum and cell culture supernatant, EV fibronectin emerged as a promising biomarker. It was able to effectively differentiate patients with NSCLC from healthy controls (AUC=0.844, P<0.001) and showed a significant increasing trend correlating with cancer progression (advanced NSCLC > early NSCLC > healthy controls, P<0.001) (61).

Tumor-educated blood platelets (TEPs)

Platelets are rich in RNA species and functional spliceosomes, undergoing specific splicing in response to the tumor microenvironment, leading to changes in RNA content. There has been an increasing focus on the role of platelets in tumorigenesis, metastasis, immune evasion and chemotherapy resistance. As tumors progress, cancer cells can educate platelets, altering their transcriptome and molecular makeup (62,63). For instance, in patients with lung cancer, TEPs exhibit significant alterations in specific RNA species, including downregulation of linc-GTF2H2-1 and RP3-466P17.2, and upregulation of lnc-ST8SIA4-12, even at early stages of the disease. Integrating TEP lincRNA, CEA, NSE and CYFRA21-1 can effectively distinguish patients with advanced-stage lung cancer from early-stage ones with an AUC of 0.899 (64).

TEPs content can be transferred via MVs. Platelets can take up tumor-derived MVs and release their own protumoral MVs, thereby establishing a blood-based network for distributing tumor-derived RNA or protein. By analyzing changes in the platelet transcriptome through sequencing and proteomics, diagnostic models based on platelet activity can be developed and used in differentiating pulmonary nodules. TEP integrin alpha 2b (TEP ITGA2B) is a promising marker for improving the identification accuracy for patients with stage I NSCLC, distinguishing malignant tumours from BPNs. TEP ITGA2B levels were significantly elevated in patients with NSCLC, with an AUC of 0.940, sensitivity of 96.4% and specificity of 81.7% for identifying stage I NSCLC from BPNs. Compared to serum CEA, TEP ITGA2B demonstrated superior performance in distinguishing benign from malignant lung nodules, particularly in the case of SPN. Further research indicated that a nomogram incorporating ITGA2B and CEA may enhance diagnostic accuracy and predict overall survival (65).

Several studies have explored the association between pulmonary nodule with platelet characteristics, such as platelet-to-lymphocyte ratio (PLR). A higher PLR has been associated with an increased risk of positive nodules and lung cancer [odds ratio=1.29 (95% CI, 0.99-1.68)] (66). To enhance diagnostic accuracy and reduce bias, a multi-index diagnostic model called Sichuan Hospital of Cancer (SCHC), incorporating platelet features (platelet counts in platelet-rich plasma samples, plateletcrit in platelet-rich plasma samples and plateletcrit in whole-blood samples), age and nodule size, has shown promising results in distinguishing benign from malignant nodules. The SCHC model outperformed other clinical models (Veterans Affairs, Mayo Clinic, Brock University) by minimizing misclassification of malignant tumors and significantly improving reclassification metrics such as net reclassification improvement and integrated discrimination improvement. This platelet-based model could aid in accurately diagnosing early-stage malignancies and guiding optimal patient management in clinical settings (67).

ncRNAs

ncRNA refers to a group of RNA molecules that do not encode proteins but play important regulatory roles in cellular processes. Epigenetic-related ncRNAs include miRNAs, small interfering (si)RNAs, PiWi-interacting RNAs and lncRNAs.

miRNAs

miRNAs are small endogenous RNAs that target mRNAs, leading to post-transcriptional silencing and potentially influencing tumor development and metastasis (68). Dysregulation of specific miRNAs or miRNA groups is closely associated with cancer progression (69). Research has explored the use of circulating miRNAs for diagnosing pulmonary nodules. miR-499a (70) and a group of other miRNAs, including miRNA-17, -146a, -200b, -182, -221, -205, -7, -21, -145 and -210 (71), were significantly elevated in NSCLC compared to BPN, and they both have potential in differentiating between benign and malignant pulmonary nodules. In addition to single miRNAs, combining multiple miRNAs can enhance the accuracy of diagnosis. A panel of miRNAs (miR-199a-3p, -148a-3p, -210-3p, -378d and -138-5p) in blood has been validated for early diagnosis of LUAD from pulmonary nodules. The use of this miRNA panel alongside CT scans significantly reduces false positives. For instance, the false-positive rate of CT imaging for nodules and ground glass nodules was reduced from 33.1 to 3.2% when positive miRNA panel results were combined with nodule size (72). In addition, miR-21 and miR-210 levels were higher, while miR-486-5p levels were lower in patients with malignant lung nodules compared to benign ones. A combination of three miRNAs achieved an AUC of 0.86 in distinguishing lung cancer from BPN with 75.00% sensitivity and 84.95% specificity (73). Furthermore, paired miRNA ratios were used in the differential diagnosis of NSCLC and BPN. Five miRNA ratios (miR-92a-3p/miR-146b-3p, miR-20a-5p/miR-146b-3p, miR-19a-3p/miR-146b-3p, miR-15b-5p/miR-146b-3p and miR-16-5p/miR-146b-3p) showed higher expression levels in NSCLC compared to BPN, with a sensitivity of 0.70 and specificity of 0.90 (74).

LncRNA

LncRNAs, which are >200 nucleotides long and lack a protein-coding function (75), play a role in malignant behavior by affecting gene transcription (76). Specifically, the concentration of lncRNA THRIL is elevated in patients with lung cancer compared to those with BPN, showing promise in distinguishing between benign and malignant nodules (77). In patients with NSCLC, lncRNA XLOC_009167 levels were significantly higher than in healthy controls or individuals with pneumonia, with an AUC of 0.740, 78.7% sensitivity and 61.8% specificity in differentiating lung cancer from healthy controls. In addition, lncRNA XLOC_009167 was able to differentiate between lung cancer and pneumonia with 90.1% sensitivity, 50.0% specificity and an AUC of 0.701 (78).

Protein functional effector RNAs (pfeRNAs)

PfeRNAs are a unique type of small ncRNA that directly binds and regulates phosphorylated proteins involved in lung cancer tumorigenesis (79). Unlike miRNAs and siRNAs, pfeRNAs enhance the function of their target proteins instead of degrading them (80). Recent research has demonstrated that pfeRNAs can serve as diagnostic markers for pulmonary nodules. An 8-pfeRNA classifier (pfeRNAa to pfeRNAh) identified through deep sequencing can effectively differentiate between malignant and BPNs, with a sensitivity of 77.1% and specificity of 74.25% (81).

ctDNA

ctDNA is fragmented DNA from tumors found in the bloodstream, typically released by necrotic or apoptotic tumor cells in the tumor microenvironment (82). Mutated or methylated ctDNA can be identified using PCR or DNA sequencing to track lung cancer development. ctDNA exhibits strong tissue specificity and correlates well with tumor tissue DNA, making it a valuable biomarker for monitoring lung cancer progression.

Tumor-specific gene mutations

Gene mutations play a crucial role in tumor development by activating oncogenes and deactivating tumor suppressor genes. Mutations in ctDNA directly reflect tumor mutations, providing a valuable tool for distinguishing between malignant and BPNs. However, the sensitivity of ctDNA may be limited by interference from wild-type sequences (83). Recent studies utilizing targeted next-generation sequencing have identified specific gene mutations in ctDNA, such as RNF213, with high specificity of 100% in distinguishing between benign and malignant pulmonary nodules (84). Plasma ctDNA shows promise in early cancer detection, particularly when combined with clinical information and traditional biomarkers. Analysis of 65 lung cancer-related genes in plasma ctDNA revealed increasing mutation levels with tumor progression, particularly in driving genes. The ctDNA assay had a sensitivity of 69% and specificity of 96% for distinguishing lung nodule nature. Combining ctDNA, serum biomarkers and patient age boosted sensitivity to 80% and specificity to 99% (85).

ctDNA methylation

DNA methylation is a well-researched epigenetic modification, particularly in gene promoter regions, often leading to tumor suppressor gene inactivation and cancer development (86). Analyzing ctDNA methylation has emerged as a novel, sensitive and non-invasive method for early lung cancer detection and distinguishing lung cancers from BPN. DNA bisulfite sequencing was conducted on 309 pulmonary nodule tissue specimens to identify cancer-specific patterns. From 3,886 hypermethylated regions in the tissue, 71 regions found in plasma were refined to select 9 markers for a diagnostic model. In a validation set, the model had 79.5% sensitivity and 85.2% specificity in distinguishing lung cancer from BPNs (87). In addition to hypermethylation changes, ctDNA hypomethylation can also aid in diagnosing pulmonary nodules. Research indicates that methylation levels of seven sites of fucosyltransferase 7 in lung cancer were significantly lower than in normal controls. Furthermore, the levels of methylation at CpG-4 and CpG-7 were lower in lung cancer compared to BPN (88).

ctDNA is typically scarce and can be surrounded by a background of DNA from healthy tissues (89). To improve sensitivity and specificity, a combination of diagnostic markers and technological innovations was employed. A study involving 246 CT-detected patients with small pulmonary nodules (diameter, ≤3.0 cm) discovered elevated methylation levels of tachykinin precursor 1 (TAC1), cysteine dioxygenase type 1 (CDO1), homeobox A7 (HOXA7) and SRY-box transcription factor 17 (SOX17) in peripheral blood in cases of NSCLC compared to benign cases. The combination of SOX17, CDO1 and HOXA7 achieved a sensitivity of 90%, specificity of 71% and an AUC of 0.88 for diagnosis (90). To further promote the clinical application of these biomarkers, a new methylation analysis technique, multiplex digital methylation-specific PCR (MSP), was developed by Zhao et al (91), which increased the sensitivity from 88 to 90% and specificity from 60 to 82% for the combination of TAC1, CDO1, HOXA7 and SOX17. Another novel testing technology (92) utilized multi-locus qPCR to screen methylation markers, selecting the highest AUC marker as the anchor. This approach, combined with 10×4-fold cross-validations for each addition, resulted in the creation of two models: LunaCAM-D with 6 methylation markers to distinguish lung cancer from benign diseases and LunaCAM-S with 5 markers to classify lung cancer from healthy controls. In a recent study, methylated DNA (prostaglandin E receptor 4, ras association domain family 1A and short stature homeobox gene 2) was combined with a radiologic feature (pulmonary nodule diameter) to create prediction models. These models achieved an AUC value of 0.948 with sensitivity and specificity of 89.5 and 95.4%, respectively, for identifying malignant from BPN (93). Another study integrated ctDNA methylation, clinical features and CT imaging features to develop a composite model named PULMOSEEK Plus. This model demonstrated a sensitivity of ≥95% across all stages of lung cancer, an AUC value of 0.98 in early-stage lung cancer and an AUC value of 0.99 in indeterminate nodules (5-10 mm). By reclassifying pulmonary nodules with two cutoffs (0.65 and 0.89), unnecessary invasive procedures could have been reduced in 85% of BPN and delayed treatment avoided in 72% of malignant nodules (94).

Circulating antibodies

Autoantibodies targeting tumor-associated antigens (TAAb) are commonly found in the preclinical phase of lung cancer (95), indicating their potential as promising noninvasive biomarkers with satisfactory sensitivity and specificity (96). Studies have highlighted the effectiveness of multi-TAAb panels, such as the Fc gamma receptor IIa, erythrocyte B membrane protein band 4.1 like 3, Nogo receptor-interacting protein-1 and S100 calcium binding protein A7 like 2 combination, in accurately distinguishing indeterminate pulmonary nodules (8-20 mm) with an AUC value of 0.78, 91.7% sensitivity and 57.1% specificity (97). In a seven-TAAb panel (tumor protein 53, protein gene product 9.5, SOX2, G antigen 7, GBU4-5, melanoma-associated antigen A1 and cancer-associated gene), sensitivity and specificity were 59.7 and 81.1% for early lung nodule differentiation. In addition, integrating these seven autoantibodies and imaging features into a machine learning model markedly increased the AUC from 0.75 to 0.96 in distinguishing patients with pulmonary nodules (98). A recent study utilized high-throughput protein microarrays to identify TAAb and developed a random forest model with six autoantibodies (annexin 2, dermcidin, MID1 interacting protein 1, paraneoplastic antigen MA1, TATA-box binding protein associated factor 10, zinc finger protein 696) to detect high-risk pulmonary nodules suitable for LDCT scans (99). In addition, another study investigated antibody responses against bacterial and viral proteins in patients with lung cancer, finding more prevalent antibodies among BPNs than among LUAD. Then a panel of 20 antibodies was created to distinguish LUAD from BPNs with an AUC of 0.80 (100).

Challenges in clinical transformation of liquid biopsy

Each liquid biopsy method has distinct advantages, limitations, diagnostic performance and cost-effectiveness profiles, necessitating biomarker selection based on clinical context (Table II). However, all of these detection approaches face technical challenges in clinical translation. For instance, ultracentrifugation remains the gold standard for EVs isolation, its limitations-including being time-consuming, low-throughput and potentially damaging to vesicles-along with protein aggregate contamination, necessitate more efficient techniques such as nanosensors (101) and microfluidic chips (102). Similarly, TEPs face challenges in specificity due to blood cell contamination during isolation (103) and an incomplete understanding of spliceosome regulation (104), requiring further multicenter validation (105). ncRNAs, despite their detectability, have shortcomings of low abundance and interference in bodily fluids, demanding improved quantification methods (106). Although ctDNA enables real-time tumor monitoring, its low concentration necessitates costly high-sensitivity assays (e.g., nanoparticle-based DNA extraction/qMSP), complicating analysis (107). Circulating antibodies, while promising, risk false positives (cross-reactivity with infections/autoimmunity) and false negatives (immunosuppression), underscoring the need for multi-analyte diagnostic panels. Collectively, advancing isolation technologies, standardizing detection and integrating multi-omics approaches are critical for robust clinical translation of liquid biopsy.

Table II Comparison of diagnostic performance and cost-effectiveness of various liquid biopsy biomarkers.

Table II

Comparison of diagnostic performance and cost-effectiveness of various liquid biopsy biomarkers.

Biomarkers	Advantages	Limitations	Dynamic monitoring capabilities	Tumor localization accuracy	Diagnostic efficiency	Cost-effectiveness
EVs	High stability and information richness; overcoming the tumor of spatial heterogeneity	Lack of standardized processes; complex separation techniques	Moderate-high	Moderate-high	Moderate	Moderate
TEPs	Rapid turnaround times; easy to standardize	Lack of large and diverse studies; the mechanism of alternative splicing remains unclear; may be contaminated with other blood cells	High	Moderate	Moderate	Moderate
ncRNA	Highly tissue/cell specific; easy to detect	Low abundance; non-specific back-ground signal interference	Low-moderate	Low-moderate	Low-moderate	Low-moderate
ctDNA	Real-time monitoring of tumor burden; allowing for early detection of recurrence	Highly degraded; low concentrations	Moderate	High	Moderate	Moderate-high
Circulating antibodies	Non-invasive; direct signaling of the immune response	Heterogeneity of immune responses; low specificity	Low	Low	Low	Low
Integrative approaches	High diagnostic efficiency	High diagnostic costs; lack of standardized processes	High	High	High	High

[i] EVs, extracellular vesicles; TEPs, tumor-educated platelets; ncRNA, non-coding RNA; ctDNA, circulating tumor DNA.

Clinical applications, integrated diagnostic strategies and cost-effectiveness of liquid biopsy

Liquid biopsy is a transformative tool in lung cancer management, offering significant clinical benefits in several clinical applications, such as early detection and screening (108), providing comprehensive molecular profiling, guiding therapy (109), identifying minimal residual disease (110), and monitoring disease progression and treatment response (111). As an important complement or alternative to traditional diagnostic methods, liquid biopsy has significantly enhanced the precision and efficiency of lung cancer diagnosis and management. In particular, the integration of liquid biopsy with current diagnostic methods, such as imaging techniques, tissue biopsy and molecular profiling, can leverage the strengths of each method to provide a comprehensive understanding of the disease (108,112-117) (Table III). The cost-effectiveness of liquid biopsy is demonstrated through multiple key aspects: Population-based screening feasibility (108), reduction of unnecessary invasive procedures (118,119), accelerated diagnostic turnaround time (120) and facilitation of personalized treatment strategies (111). These advantages collectively contribute to significant cost reduction in the clinical management pathway for pulmonary nodules and lung cancer. Although the upfront costs of liquid biopsy can be high, its potential to reduce unnecessary treatments, complications and hospitalizations makes it a cost-effective option in numerous scenarios (119) (Table III). As technology advances and costs decrease, liquid biopsy is likely to become an integral part of lung cancer care, improving outcomes for patients and optimizing healthcare resource utilization.

Table III Clinical application, combined application and cost-effectiveness of LB in diagnosis of pulmonary nodule and lung cancer.

Table III

Clinical application, combined application and cost-effectiveness of LB in diagnosis of pulmonary nodule and lung cancer.

Clinical application	Single diagnostic methods	Combined application
Early detection and screening	LDCT/PET-CT: Standard for lung cancer screening in high-risk individuals; Limitations: False positives and overdiagnosis	LDCT/PET-CT + LB (108,112): Detect genetic alterations and provide molecular evidence of malignancy even in early stages; improve the accuracy of early detection and reduces unnecessary invasive procedures
Overcoming tumor heterogeneity; comprehensive molecular profiling	TB: Gold standard for diagnosis; Limitations: Invasive, risky, provides a snapshot of the tumor's genetic profile at a specific site but may miss spatially heterogeneous mutations	TB + LB (113,114): Provide a comprehensive snapshot of the tumor's genetic landscape and improve the accuracy of molecular profiling; overcome the limitations of single-site TB, which may miss spatially heterogeneous mutations; useful when TB is contraindicated or fails to yield adequate material
Identification of targetable mutations; guiding targeted therapy and immunotherapy	TB: Identify actionable mutations for targeted therapy; assess PD-L1 expression and TMB, which are predictive of response to immune checkpoint inhibitors	TB + LB (115,116): Confirm or supplement tissue biopsy findings, enabling timely initiation of targeted therapies; provide dynamic assessment of TMB and PD-L1 status, complementing tissue-based analysis; bTMB could predict response to therapy and monitor the emergence of resistance mutations
Monitoring treatment response	CT/PET: Assess tumor size and metabolic activity during treatment; Limitations: May not detect early molecular changes	CT/PET + LB (117): Complementing imaging findings; real-time monitoring of tumor dynamics; indicate treatment efficacy or the emergence of resistance mechanisms
Post-treatment surveillance; detection of MRD	Imaging: Used for post-treatment surveillance; Limitations: May not detect microscopic residual disease	Imaging + LB (110): LB can detect MRD by identifying ctDNA in patients with no radiographic evidence; help to identify patients at risk of recurrence of disease, enabling early intervention
Cost-effectiveness of LB
• Population screening (108)
The cost-effectiveness of LB in large-scale screening programs depends on the prevalence and the ability to detect early-stage disease accurately
• Reduced invasive and unnecessary procedures (118,119)
LB is less invasive and costly, reducing repeated invasive procedures and complications, lowering associated healthcare costs and improving patient outcomes
• Faster turnaround time (120)
Liquid biopsy results are often available more quickly than tissue biopsy, enabling faster treatment decisions and reducing costs associated with delayed care
• Personalized treatment (111)
The combined approach enables more precise treatment selection, reducing the use of ineffective therapies and associated costs
• Challenges to cost-effectiveness (119)
The cost of LB tests can be high, particularly for advanced genomic profiling. However, this cost is often offset by the benefits of personalized treatment and reduced need for repeated tissue biopsies

[i] MRD, minimal residual disease; LDCT, low-dose computed tomography; PET, positron emission tomography; CT, computerized tomography; TB, tissue biopsy; TMB, tumor mutational burden; bTMB, blood-based tumor mutational burden; LB, liquid biopsy.

Conclusion and perspective

Liquid biopsy, a non-invasive and convenient diagnostic tool, is increasingly utilized in clinical practice. Circulating biomarkers provide insights into the mechanisms of lung cancer, aiding in early detection, screening, diagnosis, monitoring and treatment. However, the sensitivity and specificity of EVs, TEPs, ncRNA, ctDNA and TAAb as lung cancer biomarkers may be limited by low blood concentrations and potential interference from molecules secreted by normal cells. The heterogeneous and atypical nature of tumors necessitates a refined temporal and spatial variation map of these biomarkers, with research focusing on combining multiple biomarkers (the construction of clinical models, including imaging features and patient characteristics) or levels (the establishment of biomarkers from cells, tissue, peripheral circulation) for accurate differentiation of malignant tumour and BPN (Fig. 1). Furthermore, AI-driven analysis has the potential to efficiently analyze vast and complex datasets, thereby enabling the development of diverse and efficient diagnostic and predictive models (121). Standardization of liquid biopsy methods and critical level determination is essential for clinical application. Ultimately, liquid biopsy holds promise for enhancing early pulmonary nodule diagnosis.

Availability of data and materials

Not applicable.

Authors' contributions

YT designed this study and revised the manuscript. MP, JG and YT wrote the manuscript. TA and HC generated the table and figure. LC, MC and JL participated in the literature search and collation. All authors have read and approved the final manuscript. Data authentication is not applicable.

Ethics approval and consent to participate

Not applicable.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Abbreviations:

Acknowledgments

Not applicable.

Funding

This study was funded by the Youth Interdisciplinary Special Fund of Zhongnan Hospital of Wuhan University (grant no. ZNQNJC2022008), the Natural Science Foundation of Hubei Province (grant no. 2021CFB415) and the National Natural Science Foundation of China (grant no. 81702273).

References