Retracted: A Dynamic Adaptive Ensemble Learning Framework for Noninvasive Mild Cognitive Impairment Detection: Development and Validation Study
Post-publication record
OpenAlex flags this work as retracted, but it carries no matching Retraction Watch record in this frame.
Bibliographic record
Abstract
BACKGROUND: The prompt and accurate identification of mild cognitive impairment (MCI) is crucial for preventing its progression into more severe neurodegenerative diseases. However, current diagnostic solutions, such as biomarkers and cognitive screening tests, prove costly, time-consuming, and invasive, hindering patient compliance and the accessibility of these tests. Therefore, exploring a more cost-effective, efficient, and noninvasive method to aid clinicians in detecting MCI is necessary. OBJECTIVE: This study aims to develop an ensemble learning framework that adaptively integrates multimodal physiological data collected from wearable wristbands and digital cognitive metrics recorded on tablets, thereby improving the accuracy and practicality of MCI detection. METHODS: We recruited 843 participants aged 60 years and older from the geriatrics and neurology departments of our collaborating hospitals, who were randomly divided into a development dataset (674/843 participants) and an internal test dataset (169/843 participants) at a 4:1 ratio. In addition, 226 older adults were recruited from 3 external centers to form an external test dataset. We measured their physiological signals (eg, electrodermal activity and photoplethysmography) and digital cognitive parameters (eg, reaction time and test scores) using the clinically certified Empatica 4 wristband and a tablet cognitive screening tool. The collected data underwent rigorous preprocessing, during which features in the time, frequency, and nonlinear domains were extracted from individual physiological signals. To address the challenges (eg, the curse of dimensionality and increased model complexity) posed by high-dimensional features, we developed a dynamic adaptive feature selection optimization algorithm to identify the most impactful subset of features for classification performance. Finally, the accuracy and efficiency of the classification model were improved by optimizing the combination of base learners. RESULTS: The experimental results indicate that the proposed MCI detection framework achieved classification accuracies of 88.4%, 85.5%, and 84.5% on the development, internal test, and external test datasets, respectively. The area under the curve for the binary classification task was 0.945 (95% CI 0.903-0.986), 0.912 (95% CI 0.859-0.965), and 0.904 (95% CI 0.846-0.962) on these datasets. Furthermore, a statistical analysis of feature subsets during the iterative modeling process revealed that the decay time of skin conductance response, the percentage of continuous normal-to-normal intervals exceeding 50 milliseconds, the ratio of low-frequency to high-frequency (LF/HF) components in heart rate variability, and cognitive time features emerged as the most prevalent and effective indicators. Specifically, compared with healthy individuals, patients with MCI exhibited a longer skin conductance response decay time during cognitive testing (P<.001), a lower percentage of continuous normal-to-normal intervals exceeding 50 milliseconds (P<.001), and higher LF/HF (P<.001), accompanied by greater variability. Similarly, patients with MCI took longer to complete cognitive tests than healthy individuals (P<.001). CONCLUSIONS: The developed MCI detection framework has demonstrated exemplary performance and stability in large-scale validations. It establishes a new benchmark for noninvasive, effective early MCI detection that can be integrated into routine wearable and tablet-based assessments. Furthermore, the framework enables continuous and convenient self-screening within home or nonspecialized settings, effectively mitigating underresourced health care and geographic location constraints, making it an essential tool in the current fight against neurodegenerative diseases.
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 | 0.005 | 0.006 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".