Detection and Biopsy Guidance of Nasopharyngeal Carcinoma Based on Artificial Intelligence and Endoscopic Images

Recruiting

• Conditions: Nasopharyngeal Carcinoma; Image-Guided Biopsy

• Registration Number: NCT05202626
• Lead Sponsor: Chinese Academy of Sciences

Brief Summary

Due to the occult anatomic location of the nasopharynx and frequent presence of adenoid hyperplasia, the positive rate for nasopharyngeal carcinoma identification during biopsy is low, thus leading to delayed or missed diagnosis for nasopharyngeal carcinoma upon initial attempt. Here, we aimed to develop an artificial intelligence tool to detect nasopharyngeal malignancies and guide biopsy under endoscopic examination based on deep learning.

Detailed Description

Due to the occult anatomic location of the nasopharynx and frequent presence of adenoid hyperplasia, the positive rate for nasopharyngeal carcinoma identification during biopsy is low, thus leading to delayed or missed diagnosis for nasopharyngeal carcinoma upon initial attempt. Here, we aimed to develop an artificial intelligence tool to detect nasopharyngeal malignancies and guide biopsy under endoscopic examination based on deep learning.

Study Timeline

• Study Start: 2021-12-01
• Study First Submitted: 2022-01-09
• Study First Submitted QC: 2022-01-09
• Study First Posted: 2022-01-21
• Status Verified: 2023-10
• Last Update Submitted: 2023-10-13
• Last Update Posted: 2023-10-17
• Primary Completion: 2023-10-30
• Study Completion: 2025-06-28

Recruitment & Eligibility

• Status: RECRUITING
• Sex: All
• Target Recruitment: 100

Inclusion Criteria

• The patient was found to have a nasopharyngeal lesion through the nasopharyngeal endoscopy and the clinicans considered it necessary to perform an biopsy.
• Hemilateral lesion with limited size.

Exclusion Criteria

• Patients with nasopharyngeal cancer, oropharyngeal cancer, hypopharyngeal cancer, etc. who have already been treated.

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Primary Outcome Measures

Name	Time	Method
Aera under the receiver operating characteristic curve (AUC)	three months	AUC of an deep learning-based model in discriminating nasopharyngeal carcinoma from bengin lesion.

Secondary Outcome Measures

Name	Time	Method
Accuray	three months	The agreement between the deep learning-based model and the histopathological diagnosis of the three biopsy specimens (inside the lesion, 5-8 mm outside the lesion, and 8-10 mm outside the lesion).

Trial Locations

Locations (5)

Fujian Provincial Cancer Hospital; 🇨🇳 Fuzhou, Fujian, China

Guangdong Provincial People's Hospital; 🇨🇳 Guangzhou, Guangdong, China

Nanfang Hospital, Southern Medical University; 🇨🇳 Guangzhou, Guangdong, China

Sun Yat-Sen University Cancer Center; 🇨🇳 Guangzhou, Guangdong, China

Hainan Provincial People's Hospital; 🇨🇳 Haikou, Hainan, China