Enhancing Manifold Convexity in Deep MRI Image Clustering Using Adversarial Learning
Bibliographic record
Abstract
Till today, no cure has been found for some diseases such as cancer and early detection remains the only reliable way to increase the survival chances for patients. Recently, machine learning has been used to aid in the diagnosis of large numbers of patients, especially using medical image data such as Magnetic Resonance Imaging images. However, due to the difficulty of obtaining large and labelled datasets, clustering techniques have been favored in this field of research with particular interest in deep clustering which has proven capable of understanding and accurately grouping highly complex data. One of the most recent developments in deep clustering came in the form of the dynamic auto-encoder (DynAE), which solves a historic problem that faced autoencoders (AEs) prior. However, it suffers from feature twist, a phenomenon which causes the cluster boundaries to flatten and become coarse and more complex which in turn hinders the clustering performance. This study proposes a new model (AdversFAE) using adversarial learning to combat feature drift in the auto-encoders by utilizing adversarial constrained interpolation in the clustering phase to escape feature twist. The proposed model was tested on six datasets from the MedMNIST v2 collection of medical datasets. The results showed that the proposed model outperformed the convolutional dynamic autoencoder (ConvDynAE) in terms of accuracy and F1_score. Interestingly, AdversFAE achieved an accuracy of 85.9%, 66.9%, and 53.5% on PneumoniaMNIST, BloodMNIST, and RetinaMNIST respectively compared to $62.1 \%, 46.3 \%$, and 31.5% achieved by ConvDynAE on the same datasets. The proposed changes proved effective in improving the convexity of cluster boundaries and hence the clustering performance.
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 | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| 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 | 0.000 |
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".