CDC心脏病2020语义向量数据库

数据集获取（Figshare）

• 源站标题：CDC Heart Disease 2020 Semantic Vector Database
• DOI：10.6084/m9.figshare.31144024.v1
• 发布日期：2026-01-24
• 文件数：1 个
• 文件总体积：469.12 MB
• 源站页面：https://figshare.com/articles/dataset/CDC_Heart_Disease_2020_Semantic_Vector_Database/31144024

命令行下载

curl -L -o "faiss_heart_index.zip" "https://ndownloader.figshare.com/files/61309162"

文件清单

• faiss_heart_index.zip（469.12 MB）

文件总体积 469.12 MB（源站 API 实测）。

数据加载示例（图像类）

from PIL import Image
import glob, os

files = (glob.glob(os.path.join(path, "**", "*.png"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.jpg"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.tif"), recursive=True))
print("图像文件数:", len(files))
img = Image.open(files[0]); print("尺寸/模式:", img.size, img.mode)

# torchvision Dataset 方式:
# from torchvision import datasets
# ds = datasets.ImageFolder(path)  # 要求 子目录=类别

目录组织与标注格式以源站说明和下载后实际文件为准。

数据说明

To ensure your dataset is professionally indexed and citable for a journal paper, your Figshare description should follow the FAIR (Findable, Accessible, Interoperable, Reusable) principles.
Here is a comprehensive template you can copy and adapt.
Description
Overview
This item contains the Semantic Vector Database derived from the CDC Indicators of Heart Disease (2022 UPDATE) dataset. This database was specifically engineered to support Retrieval-Augmented Generation (RAG) and Explainable AI (XAI) research in cardiovascular informatics. It transforms over 400,000 tabular clinical records into a high-dimensional semantic space for sub-millisecond similarity search and evidence-based clinical reasoning.
Technical Specifications

• Source Dataset: CDC Behavioral Risk Factor Surveillance System (BRFSS) 2022.
• Total Records ($N$): 445,132 individual respondent profiles.
• Embedding Model: sentence-transformers/all-MiniLM-L6-v2 (Transformer-based).
• Vector Dimensions: 384.
• Vector Library: FAISS (Facebook AI Similarity Search).
• Index Type: Flat L2 (Euclidean distance) for maximum precision.
  Methodology
  The vectorization process involved three primary stages:
• Data Preprocessing: Cleaning the 2022 CDC BRFSS raw data, focusing on 40 key health variables (e.g., HadHeartAttack, SmokerStatus, BMI, CovidPos).
• Semantic Profiling: Converting categorical tabular data into structured natural language “Clinical Profiles” to preserve the context of comorbidities.
• Indexing: Generating dense vector embeddings and persisting them into a FAISS index to enable semantic retrieval by an LLM (e.g., Llama-3.1).
  Research Utility
  This database is intended for researchers working on:
• Clinical Decision Support: Identifying high-risk patient cohorts via semantic similarity.
• Public Health Trends: Analyzing correlations between COVID-19 history and heart disease within the 2022 population.
• Agentic AI: Providing a “Ground Truth” retrieval layer for AI agents to reduce hallucinations in medical queries.
  Included Files
• index.faiss: The serialized vector index containing the embeddings.
• index.pkl: The metadata mapping for the index

许可

CC BY 4.0

引用

学术使用请引用 DOI 10.6084/m9.figshare.31144024.v1（source: https://figshare.com/articles/dataset/CDC_Heart_Disease_2020_Semantic_Vector_Database/31144024）。