Deep Learning-based sbORN Diagnostic Model

Sun Yat-Sen Memorial Hospital of Sun Yat-Sen University312 enrolled

Overview

Skull-base osteonecrosis (sbORN) is a severe long-term complication of nasopharyngeal carcinoma (NPC) post radiotherapy, which significantly diminish the quality of life, increase the risk of internal carotid artery rupture, and is frequently misdiagnosed as NPC recurrence. Novel diagnostic tools are therefore clinically significant. In this study, the investigators seek to ask if a deep-learning-based model shows a significantly higher sensitivity than radiologists. With a cross-sectional design, the investigators aim to recruit 312 participants in Sun Yat-sen Memorial Hospital, Guangzhou, China that meet the eligibility criteria.

Study Type

OBSERVATIONAL

Enrollment

312

Conditions

Nasopharyngeal Carcinoma Herpesvirus 4, Human

Interventions

No Intervention: Observational CohortOTHER

No intervention is scheduled for this observational study.

Eligibility

Sex: ALLMin age: 18 Years

Medical Language ↔ Plain English

Inclusion Criteria: * Equal to or older than 18 years old. * A history of histologically confirmed nonkeratinizing undifferentiated nasopharyngeal carcinoma. * A history of radical radiotherapy at nasopharynx. * Complete remission six months post radical radiotherapy according to RECIST 1.1. * No evidence of distant metastasis upon recruitment. * Diagnosis of sbORN given by senior radiologist with 2-4 Likert scores. * Consent to biopsy awake or under general anesthesia. * Consent to perform blood tests, EBV DNA, EBV IgAs, and MRI inspection of nasopharynx and neck. * With a written consent. Exclusion Criteria: * MRI artifacts or other factors that interfere radiological diagnosis and region of interest contouring. * Suspected lesion is not confined to nasopharynx and skull-base.

Outcomes

Primary Outcomes

Area under curve of the differential diagnosis of sbORN and NPC recurrence delivered by the deep-learning-based multimodal model.

Time frame: Baseline

Area under curve of the differential diagnosis of sbORN and NPC recurrence delivered by the radiologists.

Time frame: Baseline

Secondary Outcomes

Sensitivity of the differential diagnosis of sbORN and NPC recurrence delivered by the deep-learning-based multimodal model.

Time frame: Baseline

Specificity of the differential diagnosis of sbORN and NPC recurrence delivered by the deep-learning-based multimodal model.

Time frame: Baseline

F1 score of the differential diagnosis of sbORN and NPC recurrence delivered by the deep-learning-based multimodal model.

Time frame: Baseline

Positive predictive value of the differential diagnosis of sbORN and NPC recurrence delivered by the deep-learning-based multimodal model.

Time frame: Baseline

Negative predictive value of the differential diagnosis of sbORN and NPC recurrence delivered by the deep-learning-based multimodal model.

Time frame: Baseline

Sensitivity of the differential diagnosis of sbORN and NPC recurrence delivered by the radiologists.

Time frame: Baseline

Specificity of the differential diagnosis of sbORN and NPC recurrence delivered by the radiologists.

Time frame: Baseline

F1 score of the differential diagnosis of sbORN and NPC recurrence delivered by the radiologists.

Time frame: Baseline

Positive predictive value of the differential diagnosis of sbORN and NPC recurrence delivered by the radiologists.

Time frame: Baseline

Negative predictive value of the differential diagnosis of sbORN and NPC recurrence delivered by the radiologists.

Time frame: Baseline

Dice similarity coefficient of the MRI contouring between the deep-learning-based multimodal model and the radiologists.

Time frame: Baseline

Average surface distance of the MRI contouring between the deep-learning-based multimodal model and the radiologists.

Time frame: Baseline

Deep Learning-based sbORN Diagnostic Model

Overview

Conditions

Interventions

Eligibility

Locations (1)

Outcomes

Primary Outcomes

Secondary Outcomes

Central Contacts