165MO - Early detection of HCC by routine blood based-AI

Background

Early detection of hepatocellular carcinoma (HCC) remained challenging because of the limited sensitivity of alpha-fetoprotein (aFP) test and inconclusive ultrasonography in cirrhotic liver, limiting timely curative interventions. Building on our previous research on constructing an AI classifier based on routine blood tests for HCC detection (sensitivity: 0.80, specificity: 0.81 AUROC: 0.894 at cutoff 0.43, trained by 220k patients data), our goal is to investigate whether the AI blood signature can be detected up to 1 year before clinical diagnosis.

Methods

Collection Laboratory. HCC and CLD patients were identified with ICD codes, antiviral drug history, virology test and radiology reports. Decompensated cases were excluded. Blood records (CBC, LFT, RFT, Clotting profile) within one year prior to the diagnosis of HCC were retrieved at intervals (1-3 months, 3-6 months, 6-9 months, and 9-12 months before diagnosis). The AI classifier was applied, and risk score was calculated. The test performance and the lead time created by early AI diagnosis were calculated. The AI performance was compared with AFP (using cutoff at 20 ng/mL).

Results

The cohort included 13,703 patients (3,415 HCC and 10,288 CLD patients). The screening sensitivity at above stated intervals with cutoff score of 0.43 is 61.3%, 50.1%, 44.2%, and 41.3% respectively, while ensuring specificity over 75%. The mean lead-time for detection in HCC patients is 167 days ahead of diagnosis. Routine blood test was 5 times more commonly used than aFP during the surveillance phase. The described performance of routine blood AI is superior to that of aFP, which remained under 45% in all intervals.

Conclusions

The study findings reveal that the AI routine blood signature exists up to 1 year before the clinical diagnosis of HCC and creates a meaningful window for timely early intervention to improve cure. The AI routine blood signature might advance the diagnosis of HCC in 40% of patients by one year and potentially lead to cancer mortality reduction.

Legal entity responsible for the study

Department of Mechanical and Aerospace Engineering, HKUST.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.