Development and Validation of a Deep Learning Model to Predict Distant Metastases in Nasopharyngeal Carcinoma Using Whole Slide Imaging and MRI

Development and Multicenter Validation of a Deep Learning Model Based on Whole Slide Imaging and Magnetic Resonance Imaging of the Nasopharynx and Lymph Nodes to Predict Distant Metastases at Diagnosis in Nasopharyngeal Carcinoma

Status

Recruiting

Phases

Unknown

Study type

Observational

Source

ClinicalTrials.gov

Registry ID

NCT06831357

Enrollment

500

Registered

2025-02-18

Start date

2025-02-15

Completion date

2026-12-31

Last updated

2025-02-25

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

Conditions

Distant Metastasis, Nasopharyngeal Cancinoma (NPC)

Keywords

Nasopharyngeal Cancinoma (NPC), Distant Metastasis, PET/CT, MRI, whole slide imaging, deep learning model

Brief summary

An AI model was developed to predict the likelihood of distant metastasis in patients with nasopharyngeal cancer based on pathology slides and MRI scans of the primary tumor. The model was validated using data from multiple centers. It was then applied to patients with advanced stages who were recommended to undergo PET/CT scans based on the NCCN or CSCO guidelines. This AI model can accurately screen patients with high risk of distant metastasis at the time of initial diagnosis to receive PET/CT, avoid excessive examination of patients with low risk of distant metastasis, save medical resources and reduce the economic burden on patients.

Detailed description

An AI model was constructed based on HE-stained pathological sections of the primary lesion and MRI of the nasopharynx and neck to predict the probability of distant metastasis at the first visit, and the AI model was fully verified by multicenter data; the AI model was applied to T3-4 or N2-3 patients who were recommended to undergo PET/CT examination according to the NCCN and CSCO guidelines, and the threshold of the AI model when the negative predictive value for predicting M0 was not less than 95% was determined, providing theoretical support for patients predicted by AI to be exempted from PET/CT examination.

Interventions

None listed

Study design

Observational model

COHORT

Time perspective

PROSPECTIVE

Eligibility

Sex/Gender

ALL

Healthy volunteers

Inclusion criteria

A. The primary lesion was pathologically confirmed as nasopharyngeal carcinoma (WHO classification is I, II and III); B. The stage was T3-4 or N2-3, and the nasopharynx + neck MRI plain scan and enhanced scan were performed to confirm the nasopharyngeal and cervical lymph node lesions, and PET/CT or conventional examination (chest CT plain scan + enhanced scan, upper abdominal CT or MRI plain scan + enhanced scan or abdominal color Doppler ultrasound or ultrasound angiography, and whole body bone imaging) was performed to screen for distant metastases.

Exclusion criteria

Previous history of other malignant tumors (such as other head and neck squamous cell carcinomas, thyroid cancer, breast cancer, esophageal cancer, etc.).

Design outcomes

Primary

Measure	Time frame	Description
Negative predictive value	through study completion, an average of 2 year	NPV measures the proportion of predicted negative cases that are actually negative. It tells us how reliable the model is when it predicts a negative outcome.

Secondary

Measure	Time frame	Description
Sensitivity, specificity, and positive predictive value	through study completion, an average of 2 year	Sensitivity, specificity, and positive predictive value of AI in predicting distant metastasis at the threshold corresponding to a negative predictive value of 95%.

Countries

China

Contacts

Primary ContactPu-Yun OuYang

ouyangpy@sysucc.org.cn+8618565382769

Outcome results