AI-Assisted Pathologist Performance Improvement: A Multicenter, Prospective, Randomized Controlled Trial

Artificial Intelligence Model-Assisted Improvement of Pathologists' Performance in Clinical Diagnostic Tasks: A Multicenter, Prospective, Randomized Controlled Trial

The investigators plan to conduct a multicenter, prospective, randomized controlled trial to systematically evaluate the added value of pathology-based AI models in the gastric cancer diagnostic workflow. The study will focus on comparing AI-assisted platform interpretation with conventional independent slide reading in terms of diagnostic accuracy (e.g., AUC), reading efficiency (e.g., comparison of time to diagnosis), quality of diagnostic reports, diagnostic confidence (Likert scale), and pathologists' satisfaction with the AI models. The investigators will also assess superiority for less-experienced (junior) pathologists and noninferiority for more-experienced (senior) pathologists. Successful completion of this project will provide high-level prospective evidence to support the standardized deployment, quality control, and broader application of pathology AI in the gastric cancer care pathway.

Study Overview

• Status: Enrolling by invitation
• Conditions: Pathology Foundation Model
• Intervention / Treatment: Other: AI pathology model; Other: Control

• Study Type: Interventional
• Enrollment (Estimated): 1000
• Phase: Not Applicable

Contacts and Locations

Study Locations

• China
  • Guangdong
    • Guangzhou, Guangdong, China, 510515
      • Nanfang Hospital, Southern Medical University
  • Henan
    • Zhengzhou, Henan, China
      • the First Affiliated Hospital of Zhengzhou University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

Description

Inclusion Criteria:

1. Sex: ≥ 18 years of age;
2. Patients undergoing gastric mucosal biopsy or gastric cancer surgical resection, with available digital pathology images and clinical information.

Exclusion Criteria:

1.Missing data or data of insufficient quality for analysis

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Other
• Allocation: Randomized
• Interventional Model: Crossover Assignment
• Masking: Single

Number of Arms

2

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: AI-assisted group; Doctors in this group are required to use the AI pathology diagnostic model to assist their diagnoses. The AI pathology model will provide a predicted result for each case.	Other: AI pathology model; Doctors in this group are required to use the AI pathology model to assist their diagnoses. The AI pathology model will provide a predicted result for each case.
Placebo Comparator: Independent Diagnosis Group (Control Group); In this group, pathologists will independently diagnose each case based on their own clinical experience, and will record both their time to diagnosis and their diagnostic confidence.	Other: Control; Pathologists will independently diagnose each case based on their own clinical experience, and will record both their time to diagnosis and their diagnostic confidence.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Area under ROC curve (AUC)	Area under the curve	Assessments will be conducted within one week after the physicians' diagnoses.

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Diagnostic time per case	Time required for the pathologist to complete the diagnosis of each case in the AI-assisted diagnosis group compared with the independent diagnosis group. Diagnostic time is defined as the duration (in minutes/seconds) from initiating case review to finalizing and submitting the diagnostic report in the study system	Measured immediately after the physician's diagnosis.
Diagnostic report quality score	Quality score of pathology diagnostic reports in the AI-assisted diagnosis group compared with the independent diagnosis group. Report quality will be evaluated by an independent panel of expert pathologists using a predefined scoring rubric (e.g., 0-100 scale), considering diagnostic accuracy, completeness, clarity, and structure of the report. Higher scores indicate better report quality.	Within 1 week after the initial diagnosis for each case.
Pathologists' diagnostic confidence	Self-reported diagnostic confidence of pathologists for each case in the AI-assisted diagnosis group compared with the independent diagnosis group. Diagnostic confidence will be rated by the reporting pathologist on a [5]-point Likert scale (e.g., 1 = very uncertain to 5 = very confident) immediately after completing the diagnosis. Higher scores indicate greater diagnostic confidence.	At the time of diagnosis for each case.
Pathologists' satisfaction with the AI pathology model	Overall satisfaction of pathologists with the AI pathology diagnostic model in terms of usability and perceived effectiveness. Satisfaction will be assessed using a structured questionnaire comprising Likert-scale items that evaluate ease of use, integration into workflow, clarity of AI outputs, perceived impact on diagnostic efficiency, and perceived impact on diagnostic accuracy and confidence. Higher scores indicate higher satisfaction, better usability, and greater perceived effectiveness.	Assessed once at the end of the AI-assisted reading period for each pathologist.

Collaborators and Investigators

• Sponsor: Nanfang Hospital, Southern Medical University
• Investigators: Study Director: Li Liang, Nanfang Hospital, Southern Medical University

Publications and helpful links

General Publications

• Wu Y, Qian B, Li T, Qin Y, Guan Z, Chen T, Jia Y, Zhang P, Zeng D, Moroi S, Raman R, Thinggaard BS, Pedersen F, Nehe JAO, Kamalden TA, Zhou Y, Jin Y, Li H, Ran AR, Yang D, Meng Z, Peng Q, Zheng YF, Wang D, Ji H, Zang P, Yin C, Shen J, Chen Y, Yu W, Dai R, Zhang C, Zhao X, Wang X, Chen Y, Wu Q, Xie H, Szeto SKH, Chan JYY, Chan VTT, Xie HT, Wei R, Li J, Ma W, Zhu L, Wang H, Fu H, Wang W, Lin S, Xu Z, Guan N, Zhang X, Grzybowski A, Golebiowska-Bogaj M, Gawecki M, Smedowski A, Szaraniec W, Wu Y, Wen Y, Chen X, Yao Y; EyeFM Global Reader Study Team; Lim LL, Cheung CY, Tan GSW, Grauslund J, Ruamviboonsuk P, Sivaprasad S, Keane PA, Wang YX, Tham YC, Cheng CY, Wong TY, Sheng B. An eyecare foundation model for clinical assistance: a randomized controlled trial. Nat Med. 2025 Oct;31(10):3404-3413. doi: 10.1038/s41591-025-03900-7. Epub 2025 Aug 28.

Study record dates

Study Major Dates

• Study Start (Actual): November 1, 2025
• Primary Completion (Estimated): June 1, 2027
• Study Completion (Estimated): November 1, 2027

Study Registration Dates

• First Submitted: December 5, 2025
• First Submitted That Met QC Criteria: December 5, 2025
• First Posted (Actual): December 18, 2025

Study Record Updates

• Last Update Posted (Actual): February 10, 2026
• Last Update Submitted That Met QC Criteria: February 6, 2026
• Last Verified: February 1, 2026

More Information

Terms related to this study

• Keywords: Randomized Controlled Trial
• Other Study ID Numbers: NFEC-2025-653

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No