Bibliographic record
Abstract
Background and Objective(s): Early identification of infants at high risk for abnormal neurodevelopment is critical so that interventions can begin without delay.The general movement assessment (GMA) is a validated qualitative tool shown to reliably predict the risk of motor disability including cerebral palsy (CP).However, the GMA can be resourceintensive, and maintaining certification among healthcare staff limits accessibility.We hypothesize that machine learning models such as pose estimation and keypoint detection can complement the qualitative analysis of infants' general movements (GMs) and predict differences in motor trajectory.Objective: To develop an automated quantitative method to analyze infants' GMs utilizing machine learning techniques. Study Design: Retrospective. Study Participants & Setting: Academic NICU; 76 infants.Materials/Methods: GMA videos were obtained as part of clinical care and video quality was optimized by recording videos from above and ensuring all limbs were visible at all times.A two-step model composed of (1) pose estimation and (2) CP movement pattern detection is in development.In the first step, a pre-trained neural network uses real-time pose estimation to determine body coordinates.The infant's poses are inferred by detecting 17 anatomic keypoints (eyes, ears, nose, shoulders, elbows, wrists, hips, knees, and ankles) using the Detectron2 X101-FPN model.Keypoint accuracy was determined using Object Keypoint Similarity (OKS) as a standard metric.Subsequently, 10 equally spaced frames for each of the videos averaging 4.8 minutes in length were extracted for manual annotation using the Annotation Tool.The model then produced the positional coordinates of these key points relative to the image.Results: 76 infant participants were analyzed.Participants were split into 62 training participants for developing the keypoint detection model and 14 test participants for measuring key point accuracy.A benchmark model trained on the Common Objects in Context (COCO) dataset attained a model performance of 0.83 OKS.Our new model trained on custom data improved the keypoint detection accuracy performance in the NICU setting by achieving 0.91 OKS, a 9% improvement in keypoint accuracy.Conclusions/Significance: Automated pose estimation is the first step of a two-step model to quantitatively evaluate GMs in neonates.Out-of-the-box models (COCO) fail to adapt to visual patterns in the clinical setting and re-training of the model on videos captured of NICU neonates improves the pose estimation performance.The next step will be to develop the movement model using movement features such as limb velocity and angle of appendages to classify GMs.These results show that machine learning techniques are a promising avenue to provide low-cost tools for motor dysfunctions prediction in at-risk neonates.
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 | 0.003 | 0.015 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.007 | 0.004 |
| Open science | 0.004 | 0.005 |
| Research integrity | 0.006 | 0.005 |
| Insufficient payload (model declined to judge) | 0.477 | 0.283 |
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".