Prospective Study of Intraoperative Brain Tumor Photographs for Pathological and Molecular Classification
Prospective Observational Validation of a Vision-Based Model for Integrated Histopathological and Molecular Classification of Gliomas From Intraoperative Specimen Photographs
This prospective observational cohort study will evaluate the diagnostic performance and workflow feasibility of a prespecified locked artificial intelligence model that analyzes smartphone-based natural-light gross images of fresh surgical tissue together with participant age. Approximately 400 consecutive adults undergoing initial surgery for suspected intracranial adult-type diffuse glioma will be enrolled at Beijing Tiantan Hospital, Capital Medical University, between July and September 2026. Images will be acquired before frozen-section processing according to a standardized operating procedure. Model outputs will be generated in silent mode and will not be shown to the surgical or pathology teams and will not alter diagnosis or treatment. The locked model score will be compared with the final IDH mutation reference standard derived from permanent histopathology, immunohistochemistry, and indicated molecular testing. The primary objective is to estimate patient-level discrimination for IDH-mutant versus IDH-wildtype tumors. Secondary objectives include fixed-threshold performance, precision-recall performance, calibration, end-to-end technical success, and prespecified model abstention, if an abstention rule is locked before enrollment.
研究概览
详细说明
Eligible participants will be enrolled consecutively, and a screening log will document all exclusions. Fresh ex vivo tumor tissue will be photographed before freezing. Images will be deidentified and linked through a participant-surgery-specimen-image hierarchy. The locked SNAP-IDH model will segment the tumor region of interest, permit blinded quality confirmation or editing, and combine frozen iBOT image features with age using a radial basis function support vector machine. The output is a research model score rather than a clinically calibrated absolute risk. A score of frozen yoden or will be classified as IDH-mutant. Model outputs will remain hidden from all clinical decision makers. No retraining, hyperparameter tuning, threshold selection, or recalibration will use the prospective cohort. Diagnostic accuracy will be analyzed at the participant level. Image-acquisition, segmentation, inference, and reference-standard failures will be retained in the study flow and reported separately from complete-case diagnostic performance.
研究类型
注册 (估计的)
联系人和位置
学习联系方式
- 姓名:Yinyan Wang
- 电话号码:+86 13581698953
- 邮箱:tiantanyinyan@126.com
学习地点
-
-
Beijing Municipality
-
Beijing、Beijing Municipality、中国
- 招聘中
- Beijing Tiantan Hospital
-
接触:
- Yinyan Wang
- 电话号码:+86 13581698953
- 邮箱:tiantanyinyan@126.com
-
-
参与标准
资格标准
适合学习的年龄
接受健康志愿者
有资格学习的性别
取样方法
研究人群
Consecutive adults undergoing initial surgery for suspected intracranial adult diffuse glioma at the participating facility
描述
Inclusion Criteria
- Adults aged 18 years or older at enrollment.
- A newly presenting intracranial lesion is radiologically suspected to be an adult-type diffuse glioma, and the participant is scheduled to undergo initial surgical resection or biopsy at a participating site.
- The planned index surgery includes an attempt to collect fresh tumor tissue for standardized natural-light gross-image acquisition before freezing.
- Participants must be able and willing to provide written informed consent before any study-specific procedure.
Exclusion Criteria
- Two or more radiologically distinct intracranial lesions identified on preoperative imaging.
- Radiotherapy or chemotherapy directed at the current intracranial lesion before the index surgery.
- Known recurrent disease or a previous tumor-directed surgical procedure for the current lesion.
- Primary spinal cord tumor, including a cervical spinal cord tumor.
学习计划
研究是如何设计的?
设计细节
- 观测模型:COHORT
- 时间观点:PROSPECTIVE
队列和干预
团体/队列 |
干预/治疗 |
|---|---|
|
Prospective Silent-Validation Cohort. Consecutive adults undergoing initial surgery for suspected i
|
Consecutive adults undergoing initial surgery for suspected intracranial adult-type diffuse glioma will undergo standardized research imaging of fresh surgical tissue. Locked model outputs will not influence routine surgery, frozen-section assessment, permanent pathology, molecular testing, or treatment. |
研究衡量的是什么?
主要结果指标
结果测量 |
措施说明 |
大体时间 |
|---|---|---|
|
Patient-Level Area Under the Precision-Recall Curve of the Locked SNAP-IDH Score for IDH Mutation Status
大体时间:From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
Area under the precision-recall curve (AUPRC) calculated using one locked SNAP-IDH score per patient and the final reference-standard isocitrate dehydrogenase (IDH) mutation status, with IDH-mutant status as the positive class. AUPRC ranges from 0 to 1, with higher values indicating better performance. |
From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
次要结果测量
结果测量 |
措施说明 |
大体时间 |
|---|---|---|
|
Patient-Level Sensitivity of SNAP-IDH for IDH Mutation Status at the Prespecified Threshold
大体时间:From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
Percentage of patients with reference-standard isocitrate dehydrogenase (IDH)-mutant status classified as IDH-mutant by SNAP-IDH at the prespecified, locked classification threshold. Sensitivity is calculated as true positives divided by the sum of true positives and false negatives, multiplied by 100. |
From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
|
Patient-Level Specificity of SNAP-IDH for IDH Mutation Status at the Prespecified Threshold
大体时间:From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
Percentage of patients with reference-standard isocitrate dehydrogenase (IDH)-wildtype status classified as IDH-wildtype by SNAP-IDH at the prespecified, locked classification threshold. Specificity is calculated as true negatives divided by the sum of true negatives and false positives, multiplied by 100. |
From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
|
Patient-Level Brier Score of the Locked SNAP-IDH Predicted Probability for IDH Mutation Status
大体时间:From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
Mean squared difference between each patient's locked SNAP-IDH predicted probability of isocitrate dehydrogenase (IDH) mutation and the final reference-standard status, coded as 1 for IDH-mutant and 0 for IDH-wildtype. The Brier score ranges from 0 to 1, with lower values indicating lower prediction error. |
From intraoperative image acquisition until the final IDH reference standard is available, anticipated within 14 days after surgery.
|
合作者和调查者
研究记录日期
研究主要日期
学习开始 (实际的)
初级完成 (估计的)
研究完成 (估计的)
研究注册日期
首次提交
首先提交符合 QC 标准的
首次发布 (实际的)
研究记录更新
最后更新发布 (实际的)
上次提交的符合 QC 标准的更新
最后验证
更多信息
此信息直接从 clinicaltrials.gov 网站检索,没有任何更改。如果您有任何更改、删除或更新研究详细信息的请求,请联系 register@clinicaltrials.gov. clinicaltrials.gov 上实施更改,我们的网站上也会自动更新.