Patient Cohort Characteristics for Delirium Classification from Video Features

109 video recordings from 50 participants (25 male, 25 female, median age 72) were collected between August 2020 and March 2022 at a tertiary academic center. Maanasa Mendu created this dataset to develop and validate an analytic framework for delirium classification using deep learning pose estimation. A support vector machine model achieved an average classification AUC ROC of 0.79.

Use Cases

• Training machine learning models for delirium detection based on automatically extracted video features.
• Analyzing correlations between behavioral features (eye, face, limb postures/movements) and clinical delirium status.
• Benchmarking the performance of different ML models (logistic regression, gradient boosting, SVM, random forests) on a clinical video dataset.
• Studying the feasibility of using brief, non-invasive video recordings for patient monitoring in heterogeneous clinical cohorts.

Strengths

• Dataset includes 109 videos with associated structured delirium assessments, providing a labeled clinical resource.
• Keypoint extraction was validated with high sensitivity (0.94) and specificity (0.89) against human-labeled gold standards.
• The cohort is balanced by sex (25 male, 25 female participants) and includes a specific subset of 20 videos (18%) from patients with delirium.

Limitations

• Row count is unknown, which may limit suitability assessment for large-scale model training.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is small (9.5 KB), indicating limited scope, likely containing summary statistics rather than raw video or feature data.

Provenance

Source

figshare, author Maanasa Mendu.

Collection Method

Prospective cohort study with daily structured delirium assessments and brief video recordings of consenting inpatients.

Time Range

August 2020 to March 2022

Freshness

Last updated 2026-05-29 17:30:18; freshness should be verified.

Geography

Inpatient medical or neurological ward of a tertiary academic center (location unspecified).