A new study in Translational Lung Cancer Research introduces LungCanSeek, an AI-powered blood test that detects lung cancer early using four common protein markers, offering high accuracy at a fraction of the cost of traditional imaging. Developed by SeekIn Inc., the test achieves 83.5 percent sensitivity and 90.3 percent specificity in identifying lung cancer, with 77.4 percent accuracy in classifying major subtypes including adenocarcinoma, squamous cell carcinoma, and small cell lung cancer. By enabling a two-step screening process—initial blood draw followed by low-dose CT only for positives—it slashes false positives by over tenfold and cuts costs by 2.5 times, positioning it as a viable tool for resource-limited settings worldwide.
Study Design and Methodology
The multicenter retrospective analysis, conducted by a team from prominent Chinese hospitals under the leadership of SeekIn’s founder, involved 1,814 participants, comprising 1,095 confirmed lung cancer patients and 719 controls. Researchers measured levels of four established protein tumor markers—carcinoembryonic antigen (CEA), cytokeratin-19 fragment (CYFRA 21-1), pro-gastrin-releasing peptide (ProGRP), and squamous cell carcinoma antigen (SCCA)—in blood samples via standard Roche cobas analyzers. An artificial intelligence algorithm integrated these levels with patient age and gender to generate risk scores for cancer presence and subtype.
The study evaluated performance across stages, with 67.2 percent sensitivity for early-stage (I-II) cases, underscoring its value in shifting diagnoses toward curable phases. Validation occurred in independent cohorts, confirming robustness across diverse demographics. Limitations include its retrospective nature, which calls for prospective trials to assess real-world utility, and potential variability in marker assays across labs, though the use of widely available reagents mitigates this.
Performance Metrics and Subtype Differentiation
LungCanSeek excelled in distinguishing cancer from non-cancer cases, outperforming some prior blood-based tests like the EarlyCDT-Lung autoantibody panel (41 percent sensitivity at 91 percent specificity) and DNA fragmentation assays (84 percent sensitivity at 50.9 percent specificity). For subtypes, it accurately identified adenocarcinoma (most common non-small cell form) and squamous cell carcinoma with balanced precision, while small cell cases—known for rapid progression—benefited from the test’s non-invasive speed. Early detection remains pivotal, as lung cancer claims 1.8 million lives annually, with five-year survival under 20 percent for advanced stages but exceeding 60 percent if caught early.
Cost-Effectiveness and Two-Step Strategy
A modeled two-step protocol addresses low-dose CT’s drawbacks: high radiation exposure, false-positive rates up to 25 percent, and costs averaging $300-500 per scan. LungCanSeek, at $15 per test using off-the-shelf equipment, serves as the frontline filter, referring only positives (about 10 percent of screened) for imaging. This reduces overall screening expenses to roughly $40 per person while preserving 80-90 percent of true positives, per the study’s simulations. In low- and middle-income countries, where imaging access is sparse, this could expand coverage dramatically—potentially aligning with national programs like China’s ongoing pilots and supporting global goals to cut cancer mortality by 15 percent through early intervention.
Broader Context and SeekIn’s Innovations
SeekIn, founded in 2018 in Shenzhen, China, specializes in blood-based pan-cancer detection via next-generation sequencing and AI, with applications extending to postoperative monitoring and even canine oncology. LungCanSeek builds on the firm’s multi-cancer early detection platform, which has undergone trials in top Chinese hospitals. The study aligns with surging interest in liquid biopsies: Recent U.S. advances, like Johns Hopkins‘ AI-DNA fragment test (published June 2024 in Cancer Discovery), show similar promise but at higher costs, while NCI efforts emphasize biomarker refinement for high-risk smokers.
Challenges persist, including integrating AI into routine care and addressing disparities—Black and Hispanic patients face higher late-stage diagnoses due to screening barriers. Yet, LungCanSeek’s simplicity could democratize access, echoing NLST trial gains (20 percent mortality reduction via CT) but with fewer harms. Prospective studies are underway to validate scalability, potentially revolutionizing screening in underserved regions. For details, the full publication is accessible via the journal’s site, and SeekIn’s platform outlines implementation resources.
