Phonocardiogram Classification through Averaging Transforms and Domain-Balanced Cross-Mix Approaches
Bibliographic record
Abstract
Deep learning models have recently emerged as a promising tool for the classification of phonocardiogram (PCG), but their training requires large and accurately annotated datasets. Unfortunately, such datasets are not readily available. So, datasets that are collected using different sensors and those collected in controlled and uncontrolled environments are amalgamated and used as a training dataset. This introduces heterogeneity in the dataset. Further to increase the training samples, each PCG record is segmented into cycles, and a cyclebased training model is employed with cycle labels inherited from record labels. Although these labels are accurate for normal records, they are not for abnormal ones; thus leading to mislabeling of abnormal cycles. Heterogeneity and noisy labels reduce classification accuracy because of inaccurate training. Therefore, this work proposes averaging of record cycles of PCG to address noisy labels and a novel domain-balanced crossmix (DBCM) approach to handle dataset heterogeneity. Various deep learning models such as MobileNetV2 , gMLP , and MLP-Mixer are trained to classify PCG signals as either normal or abnormal. Additionally, different 2-D transforms such as spectrogram, scalogram, MFCC, and Stockwell are considered for creating input to the deep learning models. A comprehensive comparison between the proposed method and existing methods in the literature that only address dataset heterogeneity using the domain-balanced training (DBT) approach shows that the proposed method is superior by at least 10% in terms of mean accuracy. These results underscore the effectiveness of addressing noisy labels and dataset heterogeneity and utilizing 2-D representations of PCG cycles.
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 | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| 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".