Laryngeal Disease, Nasopharyngeal Neoplasms
Conditions
Brief summary
An artificial intelligence-assisted system is trained and validated by collecting nasopharyngolaryngoscopy images from patients.
Detailed description
To address the clinical pain points of traditional nasopharyngolaryngoscopy, such as incomplete visualization, inaccurate identification, and unclear imaging, this study will retrospectively collect nasopharyngolaryngoscopy images and baseline information (including gender and age) of patients who underwent nasopharyngolaryngoscopy at participating centers for model training and validation. Deep learning algorithms will be applied to construct the model. The final clinical performance evaluation of the model will be conducted using an independent, prospectively collected test cohort.
Interventions
OTHERDiagnostic
The deep learning model is trained using the training dataset and tested with the internal validation set.
Sponsors
Ruijin Hospital
Study design
Observational model
OTHER
Time perspective
OTHER
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Age ≥ 18 years; * Underwent standard electronic nasopharyngolaryngoscopy; * Patients who underwent biopsy sampling have a clear pathological diagnosis; * Signed a written informed consent form.
Exclusion criteria
* Image quality is substandard with severe motion artifacts; * Lesion images are unclear and incomplete.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| performance of lesion detection | Within 3 months after the completion of prospective data collection | The area under the receiver operating characteristic curve (ROC-AUC) of the model for abnormal lesion detection |
| performance of anatomic site recognition | Within 3 months after the completion of prospective data collection | The average precision (AP) of the model for recognizing nasopharyngeal and laryngeal anatomic sites |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Comparison of diagnostic performance between the model and physicians | Within 3 months after the completion of prospective data collection | Differences in sensitivity, specificity, and overall accuracy between the AI model and endoscopists with different years of experience |
Countries
China
Contacts
Primary ContactBin Ye, MD PhD
Outcome results
None listed