ClinicalTrials.gg
CompletedNCT07944924IMAGEUpdated Nov 18, 2026

Endoscopic Ultrasound Predictors of Malignancy in IPMNs: Toward Personalized, Data-driven Risk Stratification

An observational study in Intraductal Papillary Mucinous Neoplasms, Pancreatic Cystic Neoplasms and IPMN, sponsored by IRCCS San Raffaele. Completed. Open to participants aged 18 Years and older. Per ClinicalTrials.gov, last updated 2026-11-18.

Sponsored by IRCCS San Raffaele · Observational

Updated Nov 18, 2026Newly registeredGo to Updates ↓

Help build ClinicalTrials.gg — short surveys, remote interviews or testing, in your language. Take part →

Study type
Observational
Model
Cohort
Time perspective
Retrospective
Enrollment
198
Ages
18 Years and older
Sex
All
01

Where this study stands

Counted across the records on this site · 2026-11-18

Among 23 studies in Pancreatic Neoplasms by IRCCS San Raffaele — 9 open to participants now, 10 under way, none of them phase 3. Pancreatic Neoplasms · IRCCS San Raffaele.

Enrollment
7th of 10
198 · median 415
Started
2015
the only one that year
Observational
10 of 23
8 open now
Registered
21st of 23
by date

Among the 23 studies in Pancreatic Neoplasms by IRCCS San Raffaele, this one is the 7th largest by enrollment of the 10 observational ones that give a count, is the only one that started in 2015, is one of 10 observational ones, and was the 21st registered.

Pancreatic Neoplasms

3,902 studies on the registry are indexed under Pancreatic Neoplasms; 1,004 are open to participants now.

Browse Pancreatic Neoplasms studies →

Open to participants now

9 of the 23 studies in Pancreatic Neoplasms by IRCCS San Raffaele are open now, in 4 countries. Recruiting first, the newest first:

4 more — sign in →

By country, each count listing its studies:

All 9 open studies →

Where they have run

Who these studies take

Of the 741 Pancreatic Neoplasms studies recruiting now, 475 (64%) set no upper age limit and 152 (21%) stop at 75 or younger; 40 take children under 18; 60 take healthy volunteers.

This study: 18 Years and older · all sexes · no healthy volunteers.

The full table on the hub →

This tab as a report to print or save as PDF → With an account.

Was this useful?
Get alerts for this landscapeDownload CSVNarrow it in search →With an account: a weekly mail of what changed among 23 studies in Pancreatic Neoplasms by IRCCS San Raffaele, or all of them as a file.
02

Study summary

Intraductal papillary mucinous neoplasms (IPMNs) are among the most clinically relevant pancreatic cystic lesions because of their potential for malignant transformation into high-grade dysplasia or invasive carcinoma. Accurate preoperative risk stratification is essential to identify patients who may benefit from surgical resection while avoiding unnecessary surgery in those with low-grade disease.

Endoscopic ultrasound (EUS) plays a central role in the evaluation of IPMNs because it allows detailed assessment of ductal and cystic features associated with malignancy, including mural nodules, main pancreatic duct abnormalities, wall thickening, contrast enhancement, and elastography patterns. However, the combined predictive value of these features has not been fully standardized.

The IMAGE study is a single-center retrospective study designed to develop and validate a machine learning-based model integrating comprehensive EUS features to predict the presence of high-grade dysplasia or invasive carcinoma in patients with surgically resected IPMNs. The study includes adult patients who underwent preoperative EUS at a tertiary referral center and had a final histological diagnosis of IPMN after surgery.

The primary objective is to assess the diagnostic performance of the IMAGE model for predicting high-grade dysplasia or invasive carcinoma. Secondary objectives include evaluation of the diagnostic contribution of individual EUS feature domains and assessment of agreement between preoperative EUS classification and final histological diagnosis.

Read the detailed description

Intraductal papillary mucinous neoplasms (IPMNs) are increasingly recognized pancreatic cystic lesions characterized by a wide spectrum of biological behavior, ranging from low-grade dysplasia to high-grade dysplasia and invasive carcinoma. Because pancreatic surgery is associated with significant morbidity and mortality, accurate preoperative risk stratification is essential to identify patients who may benefit from resection while avoiding unnecessary surgery in those with low-risk disease.

Endoscopic ultrasound (EUS) represents one of the most informative imaging modalities for the evaluation of IPMNs, allowing detailed characterization of both ductal and cystic features associated with malignant transformation. These include main pancreatic duct dilation, duct wall abnormalities, intraductal nodules, cyst size and wall thickening, mural nodules, contrast-enhancement patterns, and elastographic characteristics. In addition, EUS may allow tissue acquisition when clinically indicated. However, the interpretation of these multiple features is complex, and their combined predictive value is not fully standardized in clinical practice.

The IMAGE study (IPMN Malignancy Assessment through Grading by Endoscopic ultrasound) is a single-center retrospective study based on a prospectively maintained database from a tertiary referral center for pancreatic diseases. The study includes consecutive adult patients who underwent surgical resection for IPMN between 2015 and 2023 and who had preoperative EUS performed at the study center.

Patients are eligible if they are older than 18 years, underwent preoperative EUS at the study center, and had a final histological diagnosis of IPMN after surgery. Patients are excluded if EUS was performed at another institution, if the interval between EUS and surgery exceeded six months, or if preoperative EUS showed suspicion of infiltrating pancreatic carcinoma with a macroscopic infiltrative solid component.

For each patient, demographic, EUS, cytological (when available), and histopathological data are collected. Due to the retrospective design and variability in original reports, stored EUS images and videos are independently re-reviewed by experienced endosonographers blinded to the final histological diagnosis. Morphological, contrast-enhanced, elastographic, and cytological features are standardized by consensus, with adjudication by a third blinded reviewer in case of disagreement.

EUS variables include both duct-related and cyst-related features. Duct-related variables comprise main pancreatic duct dilation, duct wall thickening, intraductal echogenicity, hypoechogenicity of the duct wall, presence and size of pancreatic duct nodules, contrast enhancement of duct nodules, and elastographic patterns. Cyst-related variables include cyst size, cyst wall thickening, contrast enhancement of the cyst wall, presence and size of mural nodules, contrast enhancement of mural nodules, and elastographic patterns.

The primary aim of the study is to develop and validate a machine learning-based model integrating comprehensive EUS features to predict the presence of high-grade dysplasia or invasive carcinoma. The model, termed IMAGE, generates a continuous risk score representing the probability of high-grade dysplasia or invasive carcinoma based on EUS-derived variables.

The dataset is divided into training and testing cohorts using a predefined split, ensuring balanced distribution of key clinical variables. The testing cohort is not used during model development. Model performance is evaluated using discrimination metrics, including area under the receiver operating characteristic curve, as well as sensitivity and specificity at a predefined positivity threshold.

Secondary objectives include evaluating the diagnostic contribution of individual EUS feature domains (ductal, cystic, and non-ductal/non-cystic features) and assessing whether a unified model integrating all EUS variables provides incremental diagnostic performance compared with models based on isolated feature domains. Additional analyses include assessment of the agreement between preoperative EUS classification and final histological diagnosis, including IPMN subtype classification and cytology-histology concordance when tissue acquisition is available.

03

Conditions studied

  • Intraductal Papillary Mucinous Neoplasms
  • Pancreatic Cystic Neoplasms
  • IPMN
  • Pancreatic Neoplasms
  • Pancreatic Cancer

Keywords

  • Endoscopic Ultrasound
  • Artificial Intelligence
  • Machine Learning
  • IPMN
  • Risk Stratification
  • Pancreatic Cancer
  • Diagnostic Model
  • Prediction
04

Who can participate

Ages eligible
18 Years and older
Sexes eligible
All
Accepts healthy volunteers
No
Sampling method
Non-probability sample

Study population

Adult patients with histologically confirmed intraductal papillary mucinous neoplasms (IPMNs) who underwent surgical resection at a tertiary referral center between 2015 and 2023 and had preoperative endoscopic ultrasound (EUS) evaluation available for analysis.

Go through the criteria one by one — yes, no or not sure — and take the list to your doctor or the study team. Kept in this browser only.

Inclusion criteria

  • Age ≥18 years
  • Patients who underwent surgical resection for intraductal papillary mucinous neoplasm (IPMN)
  • Availability of preoperative endoscopic ultrasound (EUS) performed at the study center
  • Availability of final histological diagnosis of IPMN

Exclusion criteria

Exclusion Criteria:

  • Endoscopic ultrasound (EUS) performed at another institution
  • Interval between EUS and surgery greater than 6 months
  • Preoperative suspicion of infiltrating pancreatic carcinoma with a macroscopic infiltrative solid component on EUS
  • Missing key clinical, imaging, or histological data
05

Study design

Observational model
Cohort
Time perspective
Retrospective
Enrollment
198 participants (actual)
Patient registry
No

Groups and cohorts

  • IPMN Cohort

    Adult patients with histologically confirmed intraductal papillary mucinous neoplasms (IPMNs) who underwent surgical resection between 2015 and 2023 and had preoperative endoscopic ultrasound (EUS) performed at a tertiary referral center.

    Other: Endoscopic Ultrasound (EUS) Assessment

Interventions

  • OtherEndoscopic Ultrasound (EUS) Assessment

    Observational evaluation of endoscopic ultrasound (EUS) features, including ductal and cystic characteristics, contrast-enhanced imaging, elastography, and tissue acquisition when clinically indicated, as part of routine clinical practice. No study-specific intervention is performed.

06

What researchers measure

Primary outcomes

  1. Area Under the ROC Curve (AUROC) of the IMAGE Model for Predicting High-Grade Dysplasia or Invasive Carcinoma

    Discrimination of the IMAGE machine learning model, which generates a continuous risk score from EUS-derived variables, for predicting high-grade dysplasia or invasive carcinoma in the testing cohort, using final surgical histology as the reference standard. AUROC ranges from 0 to 1, where 0.5 indicates no discrimination and 1 indicates perfect discrimination.

    Time frame: At time of surgery (histological diagnosis within 6 months from EUS)

  2. Sensitivity of the IMAGE Model for Predicting High-Grade Dysplasia or Invasive Carcinoma

    Percentage of participants with high-grade dysplasia or invasive carcinoma at final histology who are classified as positive by the IMAGE model at a predefined positivity threshold, in the testing cohort.

    Time frame: At time of surgery (histological diagnosis within 6 months from EUS)

  3. Specificity of the IMAGE Model for Predicting High-Grade Dysplasia or Invasive Carcinoma

    ercentage of participants without high-grade dysplasia or invasive carcinoma at final histology who are classified as negative by the IMAGE model at the predefined positivity threshold, in the testing cohort.

    Time frame: At time of surgery (histological diagnosis within 6 months from EUS)

Secondary outcomes

  1. AUROC of Single-Domain EUS Models for Predicting High-Grade Dysplasia or Invasive Carcinoma

    AUROC of separate machine learning models, each built on a single EUS feature domain (ductal features, cystic features, non-ductal/non-cystic features), for predicting high-grade dysplasia or invasive carcinoma, using final surgical histology as the reference standard. One value is reported for each domain. AUROC ranges from 0 to 1, where 0.5 indicates no discrimination and 1 indicates perfect discrimination.

    Time frame: At time of surgery (histological diagnosis within 6 months from EUS)

  2. Difference in AUROC Between the Integrated IMAGE Model and Single-Domain EUS Models

    Difference in AUROC between the integrated IMAGE model (all EUS features combined) and each single-domain model (ductal, cystic, non-ductal/non-cystic) for predicting high-grade dysplasia or invasive carcinoma. Positive values indicate better discrimination by the integrated model.

    Time frame: At time of surgery (histological diagnosis within 6 months from EUS)

  3. Percentage of Participants With Concordant IPMN Morphological Subtype Between Preoperative EUS and Final Histology

    Percentage of participants in whom the IPMN morphological subtype assigned at preoperative EUS (main-duct, branch-duct, or mixed-type) matches the subtype at final histological diagnosis.

    Time frame: At time of surgery (histological diagnosis within 6 months from EUS)

  4. Percentage of Participants With Concordant Cytology and Final Histology

    Among participants who underwent EUS-guided tissue acquisition, percentage in whom the cytological diagnosis is concordant with the final histological diagnosis.

    Time frame: At time of surgery (histological diagnosis within 6 months from EUS)

07

Study locations

No study locations are listed for this record.

08

References and documents

Individual participant data

Plan to share: Undecided

No publications or documents are linked to this record.

09

Timeline

1 registry update since Sep 25, 2026
ElapsedEstimatedEstimated, not yet reached

Ran from Apr 2015 to primary completion in Dec 2025, 128 months.

How these figures are worked out, and which studies are left out of them.

Registered
Apr 7, 2026
First posted
Nov 18, 2026
StartActual
Apr 12, 2015
Primary completionActual
Dec 2025
CompletionActual
Dec 2025
Last update
Nov 18, 2026
What these dates mean
Start.
When the first participant enrolled; while estimated, when the study expects to open to recruitment.
Primary completion.
When the last participant was examined or treated to collect data for the study's main question, its primary outcome. A study that stopped early gives the date its data collection stopped.
Completion.
When the last participant's last visit took place, for every outcome, secondary ones and side effects included: the end of the study. Often later than primary completion, sometimes the same day.
Actual and Est.
Actual means the registry says the date has happened. Est. (estimated) is a date the sponsor plans or expects. It can still move, and the sponsor is expected to change it to actual once it has passed.
Registered
First appeared on the registry. No changes since
Nov 18, 2026
Show all 1 update
  1. Nov 18, 2026
    First appeared on the registry

From the registry record's own update history. This site started tracking changes on Sep 25, 2026; for anything earlier, see the record history on ClinicalTrials.gov ↗

10

Registry details

Key details

Study ID
NCT07944924
Lead sponsor
IRCCS San Raffaele
Responsible party
Paolo Giorgio Arcidiacono, MD (Adjunct professor, MD, IRCCS San Raffaele) — Principal investigator
First posted
Nov 18, 2026
Start date
Apr 12, 2015
Primary completion
Dec 2025
Completion
Dec 2025
Last update
Nov 18, 2026

Oversight

Data monitoring committee
No
FDA-regulated drug
No
FDA-regulated device
No
View the source record on ClinicalTrials.gov ↗

Not currently enrolling

This study is completed, as verified in Nov 2026. You cannot join it, but the record below documents what was studied.

Follow this study

Get an email when the registry record changes — status, dates, results — or when someone posts here.

Sign in to follow

Discussion

Questions and observations about this study, from anyone following it. Not medical advice, and not a channel to the study team — their contact details are on the registry record.

Sign in to join the discussion. Reading takes no account; posting does. You choose a display name, and a pseudonym is the default.

Nothing here yet. If you are running this trial, taking part in it, or weighing whether to, this is the place to say so.

Start the discussion