AI-Driven Cancer Diagnosis and Prediction With EHR

AI-Based Cancer Diagnosis and Prediction Using Electronic Health Records

This is a multi-center, clinical study designed to evaluate the application and effectiveness of an AI-assisted predictive model for identifying and diagnosing cancer, leveraging multimodal health data.

Study Overview

• Status: Recruiting
• Conditions: Tumor
• Intervention / Treatment: Diagnostic test: AI-Based Diagnostic and Prognostic Model

Detailed Description

Cancer diagnosis and early detection are crucial for improving patient outcomes and survival rates. Early identification of cancers and appropriate intervention can significantly impact treatment success and prognosis. In clinical practice, oncologists often need to integrate a variety of patient data-including medical history, laboratory test results, imaging data such as CT scans and MRIs, and genetic markers-to make an accurate diagnosis and develop a personalized treatment plan.

To build the foundation for our work, first phase of the project was initiated in 2023, conducting a large-scale retrospective study. This foundational phase involved analyzing comprehensive, multimodal data from approximately 1 million cancer patients. The goal was to identify key patterns and build robust preliminary models.

As precision medicine becomes increasingly important, the challenge remains to identify cancer at early stages, especially when symptoms are subtle or absent. Building on the insights from our initial analysis, the project's second phase was launched in February 2025: a prospective study. This current study aims to develop and validate an AI-assisted decision-making system by integrating multimodal data from electronic health records, imaging, laboratory results, and genetic data in a real-world clinical setting. The objective is to improve diagnostic accuracy, optimize clinical workflows, and provide more personalized treatment options for cancer patients. Ultimately, through this comprehensive, two-phase approach, this system seeks to improve early detection, guide effective treatment strategies, and enhance patient survival rates.

• Study Type: Observational
• Enrollment (Estimated): 1000000

Contacts and Locations

• Study Contact: Name: Fei Liu, MD; Phone Number: +86 13810512704; Email: liufei_2359@163.com

Study Locations

• China
  • Guangdong
    • Guangzhou, Guangdong, China
      • Recruiting
      • Sun Yat-sen Memorial Hospital, Sun Yat-sen University
      • Contact: Yunfang Yu, MD; Phone Number: +86 020-81332199; Email: yuyf9@mail.sysu.edu.cn
    • Guangzhou, Guangdong, China
      • Recruiting
      • Guangzhou Women and Children's Medical Center
      • Contact: Bingzhou Li, MD; Phone Number: +86-0756-2222569; Email: mr_jerry_99@163.com
    • Guangzhou, Guangdong, China
      • Recruiting
      • Nanfang Hospital
      • Contact: Zhuomin Li, MD; Phone Number: +86-0577-85397527; Email: chetneyli.1001@gmail.com
    • Guangzhou, Guangdong, China
      • Recruiting
      • Sun Yat-sen University Cancer Hospital
      • Contact: Yuxing Lu, MD; Phone Number: +86 13161233730; Email: yxlu0613@gmail.com
  • Sichuan
    • Chengdu, Sichuan, China
      • Recruiting
      • West China Hospital
      • Contact: Kai Wang, MD; Phone Number: +86 028-85422114; Email: wkai@stu.pku.edu.cn
  • Zhejiang
    • Wenzhou, Zhejiang, China
      • Recruiting
      • First Affiliated Hospital of Wenzhou Medical University
      • Contact: Cheng Tang, MD; Email: c249325687@163.com
    • Wenzhou, Zhejiang, China
      • Recruiting
      • Second Affiliated Hospital of Wenzhou Medical University
      • Contact: Sian Liu, MD; Phone Number: +86-0577-88002888; Email: liusan@mail3.sysu.edu.cn

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

The study population consists of individuals aged 0 to 90 years who have received care at participating study centers. Participants must have comprehensive electronic health records (EHRs) available, including medical history, laboratory test results, imaging data, and genetic information (if available). Both individuals diagnosed with cancer (including pediatric and adult cancers) and healthy individuals with no history of cancer will be included in the study to evaluate the AI-assisted model's diagnostic and predictive capabilities. The study will focus on patients with complete and documented care records from the participating centers, ensuring a diverse cohort for analysis across different age groups and cancer types.

Description

Inclusion Criteria:

1、Patients with comprehensive electronic health records (EHRs), including medical history, laboratory test results, imaging data, and genetic data (if available).

2. Individuals without severe cognitive impairments or conditions that would prevent them from providing informed consent or participating in the study.

3. Parents or guardians must provide informed consent for minors, while adult participants must provide informed consent for themselves.

Exclusion Criteria:

1. Patients with incomplete or missing key electronic health record data or insufficient follow-up data.
2. Individuals with severe cognitive disorders or other terminal illnesses that would prevent meaningful participation.
3. Pregnant women (although pediatric cancers are being considered, pregnant women would be excluded for safety reasons).

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Healthy Cohort; This group consists of individuals without any diagnosed cancer. Participants in this cohort will serve as the control group for comparison to the experimental group. No interventions or treatments will be administered to this cohort, as they represent a baseline of healthy individuals.	Diagnostic test: AI-Based Diagnostic and Prognostic Model; This intervention involves an AI system that integrates multimodal data, including patient medical history, laboratory test results, imaging data, and genetic information, to predict the risk of cancer. The system uses deep learning algorithms to provide real-time, accurate predictions, enabling early identification of cancer risks. By analyzing historical health data, the model aims to predict potential cancer developments, improving early detection and treatment outcomes.
Tumor Cohort; This group consists of individuals diagnosed with cancer, including various types. Participants in this cohort will serve as the experimental group for evaluating the effectiveness of the early prediction model in identifying cancer risks and improving diagnostic accuracy.	Diagnostic test: AI-Based Diagnostic and Prognostic Model; This intervention involves an AI system that integrates multimodal data, including patient medical history, laboratory test results, imaging data, and genetic information, to predict the risk of cancer. The system uses deep learning algorithms to provide real-time, accurate predictions, enabling early identification of cancer risks. By analyzing historical health data, the model aims to predict potential cancer developments, improving early detection and treatment outcomes.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Area Under the Curve (AUC)	AUC of the ROC curve, used to quantify diagnostic accuracy. No unit (a ratio or percentage, typically expressed as a number between 0 and 1).	1 year
F1 Score	The F1 score is the harmonic mean of precision and sensitivity (recall). It is a good measure of the model's ability to identify both true positives and minimize false positives, especially in cases where the classes are imbalanced (e.g., when the number of healthy cases is much higher than disease cases). The F1 score ranges from 0 to 1, with 1 indicating perfect precision and recall.	1 year

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Sensitivity (True Positive Rate)	Sensitivity measures how well the AI model identifies true positive cases, such as correctly diagnosing pregnant women with complications or identifying neonatal disorders.	1 year
Specificity (True Negative Rate)	Specificity measures the ability of the AI model to correctly identify cases without diseases, ensuring that healthy mothers and infants are correctly identified as negative.	1 year

Collaborators and Investigators

• Sponsor: The Eye Hospital of Wenzhou Medical University

Study record dates

Study Major Dates

• Study Start (Actual): January 19, 2025
• Primary Completion (Estimated): October 1, 2025
• Study Completion (Estimated): October 1, 2025

Study Registration Dates

• First Submitted: January 19, 2025
• First Submitted That Met QC Criteria: January 19, 2025
• First Posted (Actual): January 24, 2025

Study Record Updates

• Last Update Posted (Actual): July 30, 2025
• Last Update Submitted That Met QC Criteria: July 25, 2025
• Last Verified: July 1, 2025

More Information

Terms related to this study

• Keywords: tumor; AI-Assisted Diagnosis; Early Disease Prediction
• Additional Relevant MeSH Terms: Pathologic Processes; Disease
• Other Study ID Numbers: Cancer

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
• product manufactured in and exported from the U.S.: No