Computer-aided analysis of infant respiratory patterns
Bibliographic record
Abstract
Infants recovering from surgery and anesthesia are at risk of life-threatening Postoperative Apnea (POA). There is no way to predict which infants will experience POA, and therefore all infants with postmenstrual age (PMA) ≤ 60 weeks need to be monitored in hospital for at least 12 h postoperatively. Evidence shows a link between abnormal postoperative respiratory patterns and the occurrence of POA. Thus, study of these patterns might be useful to predict an infant's risk of POA, as well as the time when such risk abates.Comprehensive study of the postoperative respiratory patterns has been limited by two main factors. First, no representative set of respiratory data from infants at risk of POA is publicly available to investigators, and so any POA study involves a data acquisition phase that requires planning, approval, and execution. All these activities consume extensive resources and so the number of infants that can be enrolled is limited by the available budget. Second, there are no appropriate tools for the comprehensive analysis of the respiratory patterns. The most accepted method is conventional manual scoring (CMS), performed by expert scorers following guidelines from the American Academy of Sleep Medicine (AASM). CMS has several limitations: it has low intra- and inter-scorer repeatability, is labor intensive, time-consuming, and expensive. Moreover, CMS does not produce a comprehensive analysis of the respiratory patterns, but rather a list of "clinically relevant" events and the time of their occurrence. Thus, any pattern not considered "clinically relevant" by the AASM guidelines is not scored and cannot be analyzed.This thesis addresses these limitations by creating a library of infant data and developing several tools for the comprehensive analysis of infant respiratory patterns. We accomplished this in 5 stages. First, we acquired a representative dataset comprising cardiorespiratory signals from infants at risk of POA, and made these data available to the public. Second, we developed a set of tools for the efficient, repeatable, and reliable manual scoring of infant respiratory patterns. We demonstrated that use of these tools produced an analysis with high accuracy and consistency, and improved intra- and inter-scorer repeatability. Third, we developed a method based on Expectation-Maximization (EM) to combine analyses from multiple manual scorers to minimize the effects of intra- and inter-scorer variability and yield "gold standard" results with very high accuracy and consistency. These two developments improved the accuracy and repeatability of manual analysis but did not address its labor intensive, time-consuming and expensive nature. The fourth stage of this thesis addressed these limitations by automating the analysis. To do this, we developed an Automated Off-line Respiratory Event Detector (AORED) that analyses respiratory patterns by comparing metrics of respiratory behavior to thresholds to determine the presence of patterns. Optimal threshold values were selected using Receiver Operating Characteristics (ROC) analysis using manual analysis results as reference. AORED analysis agreed well with the "gold standard" manual analysis. However, its thresholds were based on the results of manual analysis so its performance will be influenced by the limitations of manual scoring. The fifth stage of the work addressed this through the development of AUREA, an Automated Unsupervised Respiratory Event Analysis system that applies unsupervised, K-means clustering to the metrics of respiratory behavior to classify the respiratory patterns. K-means requires no human intervention to work, so AUREA is completely automated and is not affected by the limitations of manual scoring. The validation results showed that AUREA had substantial accuracy (significantly higher than AORED), and almost perfect consistency.The contributions from this work will impact the study of respiratory patterns and POA in 7 ways: ...
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 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.010 | 0.002 |
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".