Multimodal Deep Learning Model Predicts Pancreatic Cancer Prognosis

Prediction of Pancreatic Cancer Prognosis Using a Multimodal Deep Learning Model Based on Intratumoral Immune Microenvironment

This study describes the development and validation of a deep learning prediction model, which extracts deep learning features from preoperative enhanced CT scans and analyzes postoperative pathological specimens of pancreatic cancer patients. The aim is to predict patient prognosis and response to chemotherapy treatment.

Study Overview

• Status: Completed
• Conditions: Pancreatic Adenocarcinoma
• Intervention / Treatment: Diagnostic test: No Interventions; Diagnostic test: No Interventions

Detailed Description

This study retrospectively collected enhanced CT scan data, pathological paraffin blocks, and clinical data from pancreatic cancer patients who underwent surgery at multiple centers between March 2013 and May 2024. The pathological paraffin blocks were stained using immunohistochemistry for prognostic immune microenvironment markers, and patients were classified based on these results. Subsequently, deep learning features were extracted from enhanced CT scans, and a multimodal prediction model was constructed using imaging features and clinical information. The model's performance was evaluated using metrics including area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity.

• Study Type: Observational
• Enrollment (Actual): 247

Contacts and Locations

Study Locations

• China
  • Zhejiang
    • Hangzhou, Zhejiang, China, 310009
      • The Second Affiliated Hospital Zhejiang University School of Medicine

Participation Criteria

Eligibility Criteria

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

• Sampling Method: Non-Probability Sample

Study Population

Pancreatic cancer patients who were undergo surgery and received adjuvant chemotherapy after surgery.

Description

Inclusion Criteria:

1. Patients with pancreatic cancer, diagnosed through pathology;
2. Patients underwent surgery and received adjuvant chemotherapy after surgery.

Exclusion Criteria:

1. Missing or inadequate quality of CT,
2. Incomplete clinical or pathological data.
3. Multiple primary malignancies;
4. History of malignancy.

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Training Cohort; Patients diagnosed with pancreatic cancer who underwent surgery and other treatments at the Second Affiliated Hospital, Zhejiang University School of Medicine	Diagnostic test: No Interventions; The high-throughput extraction of quantitative image features from medical images; Diagnostic test: No Interventions; Immunohistochemical analysis
Test Cohort; Patients diagnosed with pancreatic cancer who underwent surgery and other treatments at the Fourth Affiliated Hospital, Zhejiang University School of Medicine and Hangzhou Hosptial of Traditional Chinese Medicine	Diagnostic test: No Interventions; The high-throughput extraction of quantitative image features from medical images; Diagnostic test: No Interventions; Immunohistochemical analysis

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Performance of deep learning model	The model's performance was evaluated using metrics including area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity.	Baseline treatment

Collaborators and Investigators

• Sponsor: Second Affiliated Hospital, School of Medicine, Zhejiang University
• Collaborators: Hangzhou Hospital of Traditional Chinese Medicine; The Fourth Affiliated Hospital of Zhejiang University School of Medicine
• Investigators: Principal Investigator: Yulian Wu, PhD., Second Affiliated Hospital of Zhejiang University School of Medicine

Publications and helpful links

General Publications

• Fan Y, Du B, Pu K, Sun Y, Lv C, Hu S, Song T, Wu R, Chen Y, Tang J, Zhong Y, Bian W, Wu J, Zhang H, Ding Y, Xu H, Wu Y, Li X. Noninvasive evaluation and clinical value prediction of tumor-infiltrating neutrophil-to-T-cell ratio in pancreatic ductal adenocarcinoma. NPJ Digit Med. 2026 Jan 3. doi: 10.1038/s41746-025-02303-9. Online ahead of print.

Study record dates

Study Major Dates

• Study Start (Actual): July 5, 2024
• Primary Completion (Actual): December 15, 2024
• Study Completion (Actual): January 3, 2026

Study Registration Dates

• First Submitted: December 29, 2024
• First Submitted That Met QC Criteria: December 29, 2024
• First Posted (Actual): January 6, 2025

Study Record Updates

• Last Update Posted (Actual): January 7, 2026
• Last Update Submitted That Met QC Criteria: January 5, 2026
• Last Verified: July 1, 2024

More Information

Terms related to this study

• Keywords: prognosis; Pancreatic adenocarcinoma; Deep Learning Model
• Additional Relevant MeSH Terms: Neoplasms; Neoplasms by Histologic Type; Neoplasms, Glandular and Epithelial; Carcinoma; Adenocarcinoma
• Other Study ID Numbers: PCPAI

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