Personalization of AI-enabled PAP Therapy (CMAP®) Algorithm Shows Improvement in Prediction of Respiratory Events and Preliminary Success in Further Reduction of Therapy Pressure
Bibliographic record
Abstract
Abstract Rationale: Obstructive sleep apnea (OSA) impacts approximately 900 million people globally. Despite the efficacy of positive airway pressure (PAP) therapy in treating OSA, adherence remains around 50% after one year. It has been suggested that patient discomfort from high air pressure therapy is one of the critical barriers to adherence. Our innovation, continuous management of airway pressure (cMAP®), leverages artificial intelligence (AI) for predicting and preventing OSA events and has been shown to reduce therapy pressure by ∼20% while improving patient comfort and maintaining efficacy. However, breathing patterns tend to be diverse and can vary significantly between patients. This results in the need for large amounts of data to train generalized AI models to effectively classify breathing patterns. Building on our previous study, it was our objective to investigate the effectiveness of using a transfer-learning approach to personalize cMAP® to particular patients suffering from OSA. Methods: In a previous study, we completed a randomized crossover study including 45 patients diagnosed with mild to severe OSA who were long-term PAP device users (>1 year). Building on this study, we used a transfer-learning pipeline to train personalized versions of our cMAP® algorithm using previously captured PAP airflow data from the 45 patients. Additionally, we used data obtained in our previous study to create patient-specific test sets (comprising of seven nights of the conventional automatic (A)-PAP therapy). We assessed the effect of cMAP® personalization by analyzing changes in the models’ ability to identify OSA events (maximum F1-score) compared our general cMAP® algorithm (paired t-test). As a pilot study, we also deployed personalized cMAP® algorithms to two patients for seven nights each, monitoring changes in therapy pressure and apnea-hypopnea-index (AHI) compared to general cMAP®(unpaired t-test). Results: Personalized cMAP® significantly improved the models’ ability to identify OSA events, a 38% (p<0.001) increase, on average. The two-patient deployment resulted in a further decreases in the average therapy pressure of 1.7+/-0.5 (p<0.001) and 0.6+/-0.1 (p<0.001) cmH2O compared to previous models, while maintaining AHI in the effective treatment range of less than 5 events/hr. Conclusions: Our preliminary results demonstrate the capability to further improve cMAP® therapy through patient-specific personalization using only a limited amount of personal PAP data. We believe this represents the first step towards identifying and pheynotyping therapy on individuals and specific patient populations. Overall, these improvements may further augment the comfort benefits already attributable to cMAP® therapy.
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) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
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".