Deidentified DEXA Body Composition Dataset

This dataset contains de-identified dual-energy X-ray absorptiometry (DEXA) scan records collected at private imaging clinics across the general adult population. DEXA is the clinical gold standard for body composition measurement, and this dataset captures its full output — not summary statistics, but the complete regional breakdown that clinicians and researchers rely on. The dataset contains over 18,000 scan records spanning multiple years, covering adults aged 20–64 across a broad range of body types (BMI ~18–45, body fat 12–46%). Males and females are both represented. Many patients have undergone repeat scans, which can be linked via a de-identified PatientID field, enabling longitudinal analysis of body composition change over time. <p><br/></p> ## **Data Structure** Each row represents one scan. The 258 columns are organized into four major groups: - **Bone (BMD/BMC by region):** Bone mineral density and content for left/right arms, ribs, thoracic spine, lumbar spine, pelvis, legs, and head — plus whole-body totals with T-scores and Z-scores for osteoporosis screening benchmarks. - **Lean mass by region:** Lean tissue mass and density for the same anatomical regions, including subtotals and whole-body totals. - **Fat mass by region:** Regional fat mass and density, including android (trunk/abdominal) and gynoid (hip/thigh) zones — key for metabolic risk stratification. - **Composition indices:** Derived summary metrics including total body fat %, fat mass index (FMI), lean mass index, android/gynoid ratio, trunk/leg fat ratio, visceral adipose tissue (VAT) mass, volume, and area, subcutaneous adipose tissue (SAT) area, and estimated basal metabolic rate (BMR) via three standard equations. - **Scan images:** Each scan record is accompanied by two whole-body DEXA images — a grayscale tissue view and a false-colour fat distribution heat map. These correspond directly to the tabular measurements and enable visual inspection, model training on image data, or multimodal analysis combining pixel-level and structured composition data. <p><br/></p> ## **Use Cases** - **Metabolic disease and obesity research:** Regional fat and VAT/SAT measures are among the strongest predictors of cardiometabolic risk. This dataset supports normative modelling, risk stratification, and phenotyping at scale. - **Pharmaceutical and clinical trial design:** Establish realistic body composition benchmarks and endpoints for trials in obesity, sarcopenia, osteoporosis, and metabolic syndrome. - **Computer vision and medical image analysis:** With two whole-body DEXA images per scan, researchers can develop and validate models that estimate body composition directly from images — or benchmark image-derived predictions against the paired ground-truth measurements in the same record. - **Multimodal and foundation model development:** The combination of structured tabular data and paired scan images makes this dataset well-suited for training multimodal models that jointly reason over imaging and clinical measurements — a growing area in health AI. - **Digital health and algorithm development:** Train and validate body composition estimation models (e.g., from anthropometrics, bioimpedance, or imaging) against DEXA ground truth. - **Longitudinal change modelling:** Repeat-scan linkage enables analysis of natural body composition trajectories by age, sex, and baseline phenotype — useful for understanding intervention effects or population trends. - **Bone health and fracture risk:** Full regional BMD with T- and Z-scores supports osteoporosis research and population-level bone health analysis.

原始信息汇总

数据集名称

Deidentified DEXA Body Composition Dataset

数据类型

去标识化的双能X射线吸收测定法（DEXA）扫描记录

数据规模

• 超过18,000条扫描记录
• 覆盖多年数据
• 对象为20-64岁成年人，BMI范围约18-45，体脂率12-46%
• 包含男性和女性
• 部分患者有重复扫描记录，可通过去标识化的PatientID字段关联

数据结构

每行代表一次扫描，共258列，分为四大类：

• 骨骼（BMD/BMC按区域）：左右手臂、肋骨、胸椎、腰椎、骨盆、腿、头部，以及全身总计，包含T值和Z值
• 瘦体重（按区域）：相同解剖区域的瘦组织质量和密度，含小计和全身总计
• 脂肪质量（按区域）：区域脂肪质量和密度，包括腹部（躯干/腹部）和臀/大腿区域
• 成分指数：总体脂百分比、脂肪质量指数（FMI）、瘦体重指数、腹部/臀部比值、躯干/腿部脂肪比、内脏脂肪组织（VAT）质量/体积/面积、皮下脂肪组织（SAT）面积、估算基础代谢率（BMR）
• 扫描图像：每次扫描附带两张全身DEXA图像（灰度组织视图和假色脂肪分布热力图）

适用场景

• 代谢疾病和肥胖研究
• 制药和临床试验设计
• 计算机视觉和医学图像分析
• 多模态和基础模型开发
• 数字健康与算法开发
• 纵向变化建模
• 骨骼健康和骨折风险

业务需求

• 真实世界数据（RWD）：来自私立影像诊所的去标识化全身DEXA扫描
• 机器学习：258个结构化变量加配对全身扫描图像
• 人口健康管理：分析年龄和性别队列中的身体成分分布
• 生命科学商业化：提供临床级身体成分终点指标

更新频率

每周（Weekly）

交付方式

安全共享（Secure share）

联系方式

• 销售/支持邮箱：[e-mail]

提供方

Procurement @ Zero Design（Zero Design 旗下专业数据产品分支）

类别

AI & ML、Health and Life Sciences、Life Sciences Commercialization、Machine Learning、Population Health Management、Real World Data (RWD)