Skip to content

Analysis of the Value of AI Image Analysis Combined with DeepSeek Text Analysis in Improving Liver MR Lesion Diagnosis

Analysis of the Value of AI Image Analysis Combined with DeepSeek Text Analysis in Improving Liver MR Lesion Diagnosis

Status
Active, not recruiting
Phases
Unknown
Study type
Observational
Source
ChiCTR
Registry ID
ChiCTR2600124200
Enrollment
Unknown
Registered
2026-05-08
Start date
2026-05-10
Completion date
Unknown
Last updated
2026-05-11

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Lesions in the liver

Interventions

Gold Standard:1. Pathological diagnosis obtained via surgical resection or puncture biopsy
2. Typical benign lesions (e.g., typical hemangioma, cyst) confirmed to be stable by imaging follow-up for >= 12 months
3. Hepatocellular carcinoma (HCC) or metastatic tumors that have been comprehensively diagnosed clinically.
Index test:A total of three diagnostic methods were evaluated: manual diagnosis
Pure AI visual diagnosis
AI vision +DeepSeek text analysis and diagnosis

Sponsors

The Fifth Affiliated Hospital of Sun Yat-sen University
Lead Sponsor

Eligibility

Sex/Gender
All

Inclusion criteria

Inclusion criteria: 1. Patients who underwent liver MR plain scan and dynamic enhanced scan in our hospital; 2. The image quality is clear, there are no serious artifacts, and it meets the diagnostic requirements; 3. The lesion has a definitive gold standard diagnostic result (any one of the following criteria shall be met): (1) Pathological diagnosis obtained via surgical resection or puncture biopsy; (2) Typical benign lesions (e.g., typical hemangioma, cyst) confirmed to be stable by imaging follow-up for >= 12 months; (3) Hepatocellular carcinoma (HCC) or metastatic tumors that have been comprehensively diagnosed clinically (combining tumor markers and imaging features), and whose lesion progression during follow-up conforms to the biological behavior of malignant tumors.

Exclusion criteria

Exclusion criteria: 1. Incomplete MR image sequence (missing key sequences in plainography, arterial phase, portal phase, delayed phase, or hepatobiliary phase); 2. Poor image quality, affecting AI model recognition or physician interpretation; 3. The maximum diameter of the lesion is < 5 mm, which is beyond the sensitive range of conventional AI detection; 4. Diffuse lesions of the liver (such as diffuse liver cancer) or lesion boundaries that cannot be clearly delineated.

Design outcomes

Primary

MeasureTime frame
Sensitivity;Specificity;Accuracy;

Secondary

MeasureTime frame
AUC;Average diagnostic time per case;Report quality score;

Countries

China

Contacts

Public ContactSonglin Yang

The Fifth Affiliated Hospital of Sun Yat-sen University

jiangchuansm@qq.com+86 756 252 8320

Outcome results

None listed

Source: ChiCTR (via WHO ICTRP) · Data processed: May 16, 2026