Application and Evaluation of Vision-LSTM Model in Diagnostic Ultrasound Imaging of TI-RADS Class 4b Thyroid Nodules
Trial Snapshot
- Phase
- Not Applicable
- Status
- Completed
- Sponsor
- Enrollment
- 401
- Locations
- 1
- Primary Endpoint
- Accuracy of diagnostic models
Study Overview
Brief Summary
The aim of this study was to evaluate the performance of artificial intelligence (AI) technology in the diagnosis of thyroid nodules, specifically in the field of ultrasound image analysis. It focuses on the accuracy and clinical feasibility of the AI system based on the Vision-LSTM model in the diagnosis of TI-RADS category 4b thyroid nodules.
Study Design
- Study Type
- Observational
- Observational Model
- Cohort
- Time Perspective
- Retrospective
Eligibility Criteria
- Ages
- 20 Years to 78 Years (Adult, Older Adult)
- Sex
- All
- Accepts Healthy Volunteers
- No
Inclusion Criteria
- •(1) Patients with thyroid nodules visible on ultrasound who underwent biopsy and/or surgical resection. (2) Diagnosed as TI-RADS category 4b on the basis of preoperative ultrasound images by two sonographers with more than 5 years of experience in thyroid ultrasound diagnosis. (3) All nodules underwent puncture biopsy or surgery to obtain pathologic results.
Exclusion Criteria
- •(1)The quality of the patient's ultrasound images was poor. (2) The patient has incomplete clinical and imaging data. (3) The patient has had thyroid surgery or other treatment.
Arms & Interventions
maligant
patients with maligant thyroid masses who underwent biopsy and/or surgical resection.
benign
patients with benign thyroid masses who underwent biopsy and/or surgical resection.
Outcomes
Primary Outcomes
Accuracy of diagnostic models
Time Frame: Immediately evaluated after the diagnostic model was built
The study collected ultrasound imaging data from 401 cases of TI-RADS 4b thyroid nodules at our hospital and used this data to train and validate the Vision-LSTM model. The diagnostic results of the AI model were compared with those of junior and senior clinicians to evaluate its performance in terms of diagnostic accuracy and stability; model performance was quantified using metrics such as the area under the curve (AUC) and the precision-recall curve (PR curve).
Secondary Outcomes
No secondary outcomes reported
Investigators
Ma Zhe
Chief Physician
Qianfoshan Hospital