NICU-MM Dataset: Per-Site Accuracy for Infant Pain Assessment AI

87.3% accuracy was achieved by the PANDIA AI system on a dataset of 2,847 infants across four datasets. The dataset likely contains per-site performance metrics for a multimodal AI system designed for infant pain assessment in NICUs. It was authored by Oussama El Othmani and last updated on May 26, 2026.

Use Cases

• Benchmarking multimodal AI models for clinical tasks based on the described accuracy metrics
• Evaluating federated learning frameworks for privacy-preserving medical collaboration based on the multi-site dataset
• Analyzing the trade-off between model interpretability and performance based on the concept-bottleneck approach described

Strengths

• Performance metrics are explicitly stated, including 87.3% accuracy and 92.1% clinician acceptance rate
• The underlying AI system was evaluated on a substantial cohort of 2,847 infants
• The dataset is associated with a fully publicly available code and model repository

Limitations

• The dataset is only 5.5 KB, indicating it likely contains summary statistics rather than raw patient data
• Row count is unknown, which may limit suitability assessment
• Column-level documentation is absent; field semantics must be inferred after download

Provenance

Source

Oussama El Othmani

Collection Method

Likely derived from the evaluation of the PANDIA AI system on the NICU-MM dataset.

Freshness

Last updated 2026-05-26 17:32:16; freshness should be verified