Skip to content

A prospective study on the diagnostic accuracy of an artificial intelligence deep learning model based on multimodal imaging for predicting the preoperative staging of gastric cancer

A prospective study on the diagnostic accuracy of an artificial intelligence deep learning model based on multimodal imaging for predicting the preoperative staging of gastric cancer

Status
Active, not recruiting
Phases
Unknown
Study type
Observational
Source
ChiCTR
Registry ID
ChiCTR2600118064
Enrollment
Unknown
Registered
2026-02-02
Start date
2026-02-08
Completion date
Unknown
Last updated
2026-02-09

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

Conditions

Gastric cancer

Interventions

Gold Standard:Postoperative pathological pTNM staging
Index test:Preoperative AI cTNM staging and clinical doctor's cTNM staging

Sponsors

Ruijin Hospital Affiliated to Shanghai Jiaotong University School of Medicine
Lead Sponsor

Eligibility

Sex/Gender
All
Age
18 Years to 80 Years

Inclusion criteria

Inclusion criteria: 1. Age/Gender: 18 - 80 years old, gender not restricted. 2. Diagnosis: Confirmed as gastric adenocarcinoma through gastroscopy biopsy and pathological examination. 3. Treatment plan: After multidisciplinary discussion (MDT), a radical gastrectomy is planned, and D2 lymph node dissection is scheduled (the specific surgical method will be determined by the clinical team according to routine procedures). 4. Examination and time window: Before the radical surgery, complete the routine procedures of enhanced CT and gastroscopy at this center, and meet the following requirements: (1) Time window from imaging examination to surgery: <= 14 days (if the time window is exceeded due to clinical reasons, the reason must be recorded and can be included in descriptive/sensitivity analysis); (2) CT includes at least the portal venous phase (if the center's routine includes arterial phase/balanced phase, record the phase as usual); (3) Gastroscopy images/videos can be traced and exported in the hospital system for research analysis. 5. Image availability (minimum quality requirements): (1) CT has no severe motion artifacts, contrast agent injection failure, or missing key phases; (2) Gastroscopy images have no severe blurriness/overly bleeding that obscures the lesion and makes it impossible to interpret; (3) Image quality is judged as "assessable" by the research-predefined quality control standards: (4) Qualified CT and gastroscopy images; (5) AI successfully outputs and locks; (6) Doctors complete cTNM records; (7) Radical resection is completed and complete pathological pT/pN are obtained. 6. Ethical requirements: The subject or their legal representative signs the informed consent form and is willing to cooperate with the research process and data usage.

Exclusion criteria

Exclusion criteria: 1. Previous anti-tumor treatment: Those who received neoadjuvant chemotherapy, radiotherapy, immunotherapy or targeted therapy for this gastric cancer before the surgery. 2. Previous history of gastric surgery: Those who had undergone gastric resection, gastric bypass or other surgeries that significantly altered the anatomical structure of the stomach (which may affect imaging anatomy and AI recognition). 3. Coexisting other malignant tumors: Those who had concurrent active malignant tumors, or whose previous malignant tumors had recurred or metastasized within the past 5 years. 4. Unable to undergo radical resection: Those who were unable to undergo radical resection due to distant metastasis or extensive peritoneal dissemination discovered during the surgery (such cases generally cannot obtain comparable "post-radical resection pTNM" and will be treated as screening exclusion or unassessable). 5. Missing key data or unidentifiable reference standards: Postoperative pathological data is incomplete, making it impossible to determine pT or pN (or unable to form pTNM according to the preset standards); or key lymph node information is missing, resulting in the inability to determine the stage. 6. Unavailable or unassessable imaging data: (1) Preoperative CT or gastroscopy images cannot be obtained, lost, or failed to be de-identified; (2) The imaging quality is severely substandard, and is judged as "unassessable" according to the preset quality control standards. 7. Other situations considered inappropriate for inclusion by the researchers: Such as situations that affect compliance or data integrity.

Design outcomes

Primary

MeasureTime frame
Accuracy rate;

Secondary

MeasureTime frame
Sensitivity;Specificity;Positive predictive value;Negative predictive value;F1 value;Area under the ROC curve;Consistency;Stability;

Countries

China

Contacts

Public ContactChao Yan

Ruijin Hospital Affiliated to Shanghai Jiaotong University School of Medicine

yanchaosuper@163.com+86 136 8174 9682

Outcome results

None listed

Source: ChiCTR (via WHO ICTRP) · Data processed: Feb 15, 2026