Healthcare — Early Tumor Detection 医疗健康 — 肿瘤早期风险预测

This demo illustrates how patient risk can change over time and how early warning signals for malignancy may emerge before diagnosis. By tracking monthly changes in clinical markers, the model highlights patterns that tend to rise or fall as cancer risk increases. 本示例展示患者风险如何随时间变化，并说明恶性风险在正式确诊前可能出现的提前信号。模型通过按月跟踪临床指标的变化，识别那些随着风险上升而呈现上升或下降趋势的特征。

What this demo does 示例说明

• Goal: identify rising malignancy risk early so clinicians can prioritise follow-up and further testing.
• Data: monthly synthetic patient records with 13 anonymised clinical features.
• Approach: the model examines short-term and longer-term prediction horizons to see how risk evolves.
• Alert rule (demo): an alert is raised when risk stays above a threshold for two consecutive periods.
• 目标：及早识别恶性肿瘤风险，便于医生提前安排复查与进一步检查。
• 数据：按月生成的模拟患者数据，包含 13 个匿名化临床特征。
• 方法：模型基于不同预测期限观察风险随时间的上升趋势。
• 预警规则（示例）：当风险连续两个时期高于阈值时触发预警。

Dataset overview 数据集说明

This demo uses a fully synthetic patient-level dataset created to mimic the kinds of clinical patterns seen in early tumor/cancer risk monitoring. Each patient is tracked monthly with 13 anonymised features that capture changes in biomarkers, vital signs and lifestyle factors. The goal is to illustrate how risk can evolve over time and how early warning signals may emerge before diagnosis. The dataset is standardized and split into a training set (80%) and a test set (20%), allowing the model’s behaviour to be demonstrated without using any real patient information.

本示例使用完全模拟的患者数据集，旨在仿真早期肿瘤风险监测中常见的临床变化趋势。每位患者按月记录 13 项匿名化特征，涵盖生物标志物、生命体征以及生活方式等指标，用于展示风险如何随时间演变，以及在确诊之前可能出现的提前信号。数据按照 80% 训练集、20% 测试集进行划分并正规化，整个演示过程不涉及任何真实患者信息。

Variables 变量说明

Key definitions 关键术语解释

Target distribution 标签分布

What you’ll see 示例输出内容

• How each clinical variable’s influence changes across different prediction horizons, shown through an easy-to-read significance table and interactive coefficient plots.
• A patient-level risk curve that illustrates how risk builds over time, including the alert point and the eventual diagnosis period.
• 通过显著性统计表和交互式系数图，展示不同预测期限下各临床特征影响力的变化。
• 患者层面的风险曲线，展示风险随时间的上升，以及预警时点和最终确诊阶段。

Run the demo 运行示例