PP259 Topic: AS25–Sedation/Analgesia/Delirium/Withdrawal Syndrome/Other: PSYCHOMOTOR AGITATION DETECTION USING DEEP LEARNING
Bibliographic record
Abstract
Aims & Objectives: Strained healthcare systems can benefit from AI-assisted monitoring that reduces the need of monitoring by caregivers and potentially enables earlier interventions by picking up on subtle cues. One challenging task in pediatric intensive care units is to detect psychomotor agitation in infants often manifested as restlessness or anxiety. Here, we propose an approach that uses deep learning to visually analyze motion patterns of infants. Methods: Our method consists of three steps: First, we detect skeletal structure using the object detection framework Detectron2(Wu et al.) which extracts keypoints representing infants’ limbs. Afterwards, we predict detailed motion information for each limb using FlowFormer(Huang et al.), a transformer-based architecture designed for optical flow estimation. Finally, we determine an empirical threshold to identify periods of psychomotor agitation. We adapt several heuristics tailored to our task, e.g. to distinguish infants vs adults we use the histogram over the variance across 17 keypoints associated with individuals captured within a single frame of a video (Figure 1). A set of keypoints corresponds to an infant if its variance belongs to the most populated bin in the histogram since the infant is typically in view for most of the video.Results: Our approach was evaluated on 98 videos from CHUSJ Hospital ICU in Montreal, achieving 38% precision @ 75% recall and F1 score of 0.63. Conclusions: Our initial study yields false positives as a result of our lenient approach in flagging agitated children. Our ongoing efforts involve minimizing this issue by labeling additional data to improve the accuracy. Keywords: psychomotor agitation detection, deep learning, intensive care unit
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.027 | 0.014 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".