神经系统疾病癫痫诊疗模型的微调研究

注册号:

Registration number:

ChiCTR2600121828 

最近更新日期:

Date of Last Refreshed on:

2026-04-03 15:14:24 

注册时间:

Date of Registration:

2026-04-03 00:00:00 

注册号状态:

补注册

Registration Status:

Retrospective registration

注册题目:

神经系统疾病癫痫诊疗模型的微调研究

Public title:

Fine-tuning Study of Epilepsy Diagnosis and Treatment Models for Neurological Diseases

注册题目简写:

English Acronym:

研究课题的正式科学名称:

神经系统疾病癫痫诊疗模型的微调研究

Scientific title:

Fine-tuning Study of Epilepsy Diagnosis and Treatment Models for Neurological Diseases

研究课题代号(代码):

Study subject ID:

在二级注册机构或其它机构的注册号:

The registration number of the Partner Registry or other register:

申请注册联系人:

林一聪 

研究负责人:

林一聪 

Applicant:

Lin Yichong 

Study leader:

Lin Yichong 

申请注册联系人电话:

Applicant telephone:

+86 13671086598

研究负责人电话:

Study leader's
telephone:

+86 10 83198815

申请注册联系人传真 :

Applicant Fax:

研究负责人传真:

Study leader's fax:

申请注册联系人电子邮件:

Applicant E-mail:

linyc_1@163.com

研究负责人电子邮件:

Study leader's E-mail:

linyc_1@163.com

申请单位网址(自愿提供):

Applicant website(voluntary supply):

研究负责人网址(自愿提供):

Study leader's website(voluntary supply):

申请注册联系人通讯地址:

中国北京市西城区长椿街45号

研究负责人通讯地址:

中国北京市西城区长椿街45号

Applicant address:

45 Changchunjie Street, Xicheng District, Beijing, China

Study leader's address:

45 Changchunjie Street, Xicheng District, Beijing, China

申请注册联系人邮政编码:

Applicant postcode:

研究负责人邮政编码:

Study leader's postcode:

申请人所在单位:

首都医科大学宣武医院

Applicant's institution:

Xuanwu Hospital, Capital Medical University

研究负责人所在单位:

首都医科大学宣武医院

Affiliation of the Leader:

Xuanwu Hospital Capital Medical University

是否获伦理委员会批准:

Approved by ethic committee:

Yes

伦理委员会批件文号:

Approved No. of ethic committee:

临研审[2026]037号-002

伦理委员会批件附件:

Approved file of Ethical Committee:

查看附件View

批准本研究的伦理委员会名称:

首都医科大学宣武医院伦理委员会

Name of the ethic committee:

Ethics Committee of Xuanwu Hospital, Capital Medical University

伦理委员会批准日期:

Date of approved by ethic committee:

2026-02-14 00:00:00

伦理委员会联系人:

张卓然

Contact Name of the ethic committee:

Zhang Zhuoran

伦理委员会联系地址:

中国北京市西城区长椿街45号

Contact Address of the ethic committee:

45 Changchunjie Street, Xicheng District, Beijing, China

伦理委员会联系人电话:

Contact phone of the ethic committee:

+86 10 83199270

伦理委员会联系人邮箱:

Contact email of the ethic committee:

xwzhuoranzhang@163.com

研究实施负责(组长)单位:

首都医科大学宣武医院

Primary sponsor:

Xuanwu Hospital Capital Medical University

研究实施负责(组长)单位地址:

中国北京市西城区长椿街45号

Primary sponsor's address:

45 Changchunjie Street, Xicheng District, Beijing, China

试验主办单位(项目批准或申办者):

Secondary sponsor:

国家:

中国

省(直辖市):

北京

市(区县):

Country:

China

Province:

Beijing

City:

单位(医院):

首都医科大学宣武医院

具体地址:

中国北京市西城区长椿街45号

Institution
hospital:

Xuanwu Hospital Capital Medical University

Address:

45 Changchunjie Street, Xicheng District, Beijing, China

经费或物资来源:

自筹

Source(s) of funding:

Self-funded

研究疾病:

癫痫  

Target disease:

Epilepsy

研究疾病代码:

Target disease code:

研究类型:

诊断试验

Study type:

Diagnostic test

研究所处阶段:

其它 

Study phase:

N/A

研究设计:

诊断试验诊断准确性 

Study design:

Diagnostic test for accuracy 

研究目的:

本课题通过先进的大模型微调技术,结合癫痫医疗大数据,由临床医生全程参与,在实际医疗场景中对医疗基座模型进行微调和反馈强化学习,构建一个涵盖医学知识、临床实践和诊疗规范深度融合的微调平台,最终形成面向癫痫领域的垂直医疗大模型,并进行多中心验证。  

Objectives of Study:

This study utilizes advanced fine-tuning techniques for large language models (LLMs), combined with extensive epilepsy medical big data and continuous clinician involvement. Through fine-tuning and reinforcement learning with human feedback (RLHF) on the medical foundation model in real clinical scenarios, we build a specialized fine-tuning platform that deeply fuses medical knowledge, clinical practice, and standardized diagnostic and treatment protocols. The result is a vertical epilepsy-specific large medical model, which is then validated across multiple centers.

药物成份或治疗方案详述:

 

Description for medicine or protocol of treatment in detail:

 

纳入标准:

Inclusion criteria

排除标准:

1.回顾性研究排除标准:(1)因病历修改或者涂抹无法正确读取信息;(2)检查报告涂抹或毁损导致大模型难以识别;(3)合并严重精神疾病的患者(如抑郁症、躁狂症、癫痫发作、精神分裂症等;(4)临床信息严重缺失。
2.前瞻性研究排除标准:(1)合并严重精神疾病的患者(如抑郁症、躁狂症、癫痫发作、精神分裂症等);(2)研究者认为不能配合完成试验的患者;(3)因患者口音或录音效果不佳无法正确读取信息;(4)检查报告涂抹或毁损导致大模型难以识别。

Exclusion criteria:

1. Exclusion criteria for the retrospective study: (1) Information cannot be correctly read due to modification or obliteration of medical records; (2) Examination reports are obliterated or damaged, making it difficult for the large language model to recognize them; (3) Patients with severe mental disorders (e.g., depression, mania, epileptic seizures, schizophrenia, etc.); (4) Severe lack of clinical information. 2. Exclusion criteria for the prospective study: (1) Patients with severe mental disorders (e.g., depression, mania, epileptic seizures, schizophrenia, etc.); (2) Patients deemed by the investigator as unable to cooperate in completing the trial; (3) Information cannot be correctly read due to patient accent or poor recording quality; (4) Examination reports are obliterated or damaged, making it difficult for the large language model to recognize them.

研究实施时间:

Study execute time:

From 2025-12-01 00:00:00 To 2028-12-31 00:00:00  

征募观察对象时间:

Recruiting time:

From 2026-03-09 00:00:00 To 2028-12-31 00:00:00

诊断试验:

Diagnostic Tests:

金标准或参考标准(即可准确诊断某疾病的单项方法或多项联合方法,在本研究中用于诊断是否有该病的临床参考标准):

根据2017年ILAE诊断标准明确诊断为癫痫

Gold Standard or Reference Standard (The clinical reference standards required to establish the presence or absence of the target condition in the tested population in present study):

Diagnosed with epilepsy according to the 2017 ILAE diagnostic criteria

指标试验(即本研究的待评估诊断试验,无论为方法、生物标志物或设备,均请列出名称):

在内部验证集上进行交叉验证,在外部多中心队列上进行前瞻性验证

Index test:

Cross-validation was performed on the internal validation set, and prospective validation was conducted on the external multi-center cohort.

目标人群(可以是某种疾病患者或正常人群,详细描述其疾病特征,注意应纳入符合分布特点的全序列病例,具有良好的代表性)

经癫痫专科医生确诊,符合ILAE 2014实用临床定义的癫痫(至少两次非诱发性痫性发作间隔>24小时,或一次发作伴≥60%复发风险,或明确癫痫综合征)

例数:

Sample size:

2350

Target condition (The target condition is a particular disease or disease stage that the index test will be intended to identify. Please specify the characteristics in detail; the population should has a complete spectrum and good representative):

Confirmed diagnosis of epilepsy by an epilepsy specialist, meeting the ILAE 2014 practical clinical definition of epilepsy (at least two unprovoked epileptic seizures occurring more than 24 hours apart, or one unprovoked seizure with a ≥60% risk of recurrence over the next 10 years, or a definite ep

容易混淆的疾病人群(即与目标疾病不易区分的一种或多种不同疾病,应避免采用正常人群对照的病例-对照设计):

例数:

Sample size:

0

Population with condition difficult to distinguish from the target condition, the normal population in a case-control study design should be avoid:

None

研究实施地点:

Countries of recruitment and research settings:

国家:

中国

省(直辖市):

北京 

市(区县):

 

Country:

China

Province:

Beijing

City:

单位(医院):

首都医科大学宣武医院 

单位级别:

三级甲等 

Institution
hospital:

Xuanwu Hospital Capital Medical University

Level of the institution:

Tertiary A

国家:

中国

省(直辖市):

河北 

市(区县):

 

Country:

China

Province:

Hebei

City:

单位(医院):

雄安宣武医院 

单位级别:

三级甲等 

Institution
hospital:

Xiongan Xuanwu Hospital

Level of the institution:

Tertiary A

国家:

中国

省(直辖市):

河北 

市(区县):

 

Country:

China

Province:

Hebei

City:

单位(医院):

河北医科大学第一医院 

单位级别:

三级甲等 

Institution
hospital:

The First Hospital of Hebei Medical University

Level of the institution:

Tertiary A

国家:

中国

省(直辖市):

内蒙古自治区 

市(区县):

 

Country:

China

Province:

Inner Mongolia Autonomous Region

City:

单位(医院):

赤峰市医院 

单位级别:

三级甲等 

Institution
hospital:

Chifeng Municipal Hospital

Level of the institution:

Tertiary A

测量指标:

Outcomes:

指标中文名:

每个病例诊疗任务的得分百分比(范围:0-100分)

指标类型:

主要指标

Outcome:

Percentage score for each case's diagnostic and treatment task (range: 0–100 points)

Type:

Primary indicator

测量时间点:

每例临床验证病例使用系统后24小时内统计

测量方法:

列出前三位最可能的诊断:每提出一个“合理(plausible)的诊断”得1分。 支持性发现与反对性发现:分别列出支持和反对该诊断的病史、体征或检查结果。部分正确:1分,完全正确:2分 最终诊断(Topdiagnosis):合理诊断:1分最正确诊断:2分 后续诊疗步骤:最多提出3个进一步检查/处理的下一步措施。部分正确:1分完全正确:2分

Measure time point of outcome:

Statistics were collected within 24 hours after each clinically validated case used the system.

Measure method:

List the top three most likely diagnoses: 1 point is awarded for each plausible diagnosis proposed. Supporting and opposing findings: List the history, signs, physical examination findings, or test results that support or oppose each proposed diagnosis. Partial correctness: 1 point; complete correctness: 2 points.Top (final) diagnosis: 1 point for a plausible diagnosis; 2 points for the most accurate/correct diagnosis. Subsequent diagnostic and treatment steps: Propose up to 3 next steps for fur

指标中文名:

医生对系统使用主观满意度评分(范围:0-100分)

指标类型:

次要指标

Outcome:

Physician subjective satisfaction score with system use (range: 0–100 points)

Type:

Secondary indicator

测量时间点:

每例临床验证病例使用系统后24小时内填写问卷

测量方法:

医生对系统使用的主观满意度评分;采用5点Likert量表(1=非常不满意,5=非常满意),评估整体满意度、使用便利性、诊断辅助价值等维度

Measure time point of outcome:

The questionnaire was completed within 24 hours after each clinical validation case used the system.

Measure method:

Physician subjective satisfaction score with system use; assessed using a 5-point Likert scale (1 = very dissatisfied, 5 = very satisfied), evaluating dimensions including overall satisfaction, ease of use, diagnostic assistance value, etc.

指标中文名:

癫痫病历生成任务上85%的准确率

指标类型:

次要指标

Outcome:

85% accuracy on the epilepsy medical record generation task

Type:

Secondary indicator

测量时间点:

每例临床验证病例使用系统后24小时内统计

测量方法:

现病史,既往史、用药史、过敏史等条目。完全错误或严重遗漏(致命错误)0分,部分正确/轻度错误1分,完全正确2分。

Measure time point of outcome:

Statistics were collected within 24 hours after each clinical validation case used the system.

Measure method:

Current medical history, past medical history, medication history, allergy history, and other relevant items. Scoring: Complete error or severe omission (critical/fatal error) = 0 points; partial correctness / minor error = 1 point; complete correctness = 2 points.

采集人体标本:

Collecting sample(s)
from participants:

标本中文名:

组织:

Sample Name:

NA

Tissue:

人体标本去向

其它  

说明

Fate of sample:

0thers  

Note:

征募研究对象情况:

Recruiting status:

正在进行

Recruiting

年龄范围:

Participant age:

最小 Min age 14 years
最大 Max age NA years

性别:

男女均可

Gender:

Both

随机方法(请说明由何人用什么方法产生随机序列):

Randomization Procedure (please state who generates the random number sequence and by what method):

None

是否公开试验完成后的统计结果:

Calculated Results after the Study Completed public access:

不公开/Private

盲法:

Blinding:

None

是否共享原始数据:

IPD sharing

否No

共享原始数据的方式(说明:请填入公开原始数据日期和方式,如采用网络平台,需填该网络平台名称和网址):

The way of sharing IPD”(include metadata and protocol, If use web-based public database, please provide the url):

None

数据采集和管理(说明:数据采集和管理由两部分组成,一为病例记录表(Case Record Form, CRF),二为电子采集和管理系统(Electronic Data Capture, EDC),如ResMan即为一种基于互联网的EDC:

采集:采用回顾性电子病历+前瞻性临床数据采集方式。回顾性数据从医院HIS系统提取脱敏后匿名化数据;前瞻性数据通过医院HIS系统提取脱敏后匿名化数据,语音转病历系统 管理:建立数据管理计划,包括数据字典、变量定义、编码规范。双人独立录入(一人录入、一人核对),逻辑校验规则自动运行,异常值人工复核。质量控制:定期抽样稽查(10%数据)、一致性检查、缺失值处理(多重插补或排除)。数据存储:加密服务器,访问日志记录,备份周期每周。

Data collection and Management (A standard data collection and management system include a CRF and an electronic data capture:

Data Collection: Retrospective electronic medical records combined with prospective clinical data collection. Retrospective data were extracted from the hospital HIS system, followed by de-identification and anonymization. Prospective data were extracted from the hospital HIS system with de-identification and anonymization, supplemented by a voice-to-medical-record transcription system. Data Management: A data management plan (DMP) was established, including a data dictionary, variable definitions, and coding conventions. Double independent data entry was implemented (one person entering, another verifying), with automated logical validation rules and manual review of outliers. Quality control measures included regular sampling audits (10% of data), consistency checks, and handling of missing values (multiple imputation or exclusion). Data storage was on an encrypted server with access logging and weekly backups.

数据与安全监察委员会:

Data and Safety Monitoring Committee:

无/No

注册人:

Name of Registration:

 2026-04-03 15:14:17