Dynamic risk from routine blood tests: An artificial intelligence two-point vector score for gastric cancer screening

Background:
Gastric cancer (GC) is a leading cause of cancer mortality globally, often diagnosed late with poor prognosis. Routine blood tests (RBT), including complete blood count, liver, and renal function, provide accessible data for risk assessment. Our AI-RBT-GC risk score leverages these tests for early GC detection and showed promising results. This study examines the AI-RBT-GC score dynamics up to 12 months before clinical GC diagnosis to guide early detection strategies.
Methods:
We analysed data from 100 patients (50 GC cases, 50 controls) at the Hong Kong Hospital Authority Data Collaboration Laboratory (2000-2020). For each patient, we extracted clinical history and one RBT result per interval (0-1, 1-3, 3-6, 6-9, 9-12 months before diagnosis or OGD). AI-RBT-GC scores (0-1) were calculated, with cutoffs at 0.33 (low/intermediate risk) and 0.64 (intermediate/high risk). Sensitivity was assessed using 1) single readings and 2) composite readings of two consecutive
Results:
Among 50 GC patients (median age 75 years, 56% male, median overall survival 14.9 months), the median score rose from 0.37 at 12 months to 0.70 at diagnosis. Single-reading sensitivity increased from 0.54 (27/50) at 12 months, to 0.68 (34/50) at 9 months, 0.54 (27/50) at 6 months, 0.66 (33/50) at 3 months, and 0.84 (42/50) at diagnosis. This upward trend reflects increased GC expression as the disease progresses. Composite readings (two consecutive scores >= 0.33) reduced sensitivity (0.48 at
Conclusions:
The AI-RBT-GC score shows a rising trend toward diagnosis, indicating disease progression. These findings highlight its potential to aid clinicians with patient selection for endoscopy and imaging to enhance early GC detection.