GAN-based OCT Image Quality Enhancement: Mapping from Low Quality Cirrus OCT to High Quality EDI OCT
Bibliographic record
Abstract
Optical Coherence Tomography (OCT) has emerged as a powerful imaging modality for diagnosing various retinal diseases including age-related macular degeneration (AMD), and central serous chorioretinopathy (CSCR). However, the quality of OCT scans obtained from different devices may vary significantly, affecting the accuracy of clinical analysis and decision-making. In particular, first-generation spectral domain OCT devices such as Cirrus 5000 OCT device produce poor quality OCT images while state-of-the-art devices such as enhanced depth imaging (EDI) OCT produce high-resolution scans where the retinal and choroidal structures can be visualized clearly. Accordingly, the disease diagnosis and quantification made based on the Cirrus OCT images are susceptible to errors. Specifically, clinicians have to pay a lot of attention to an OCT B-scan to understand the structural changes. However, over the last two decades, the low-quality Cirrus 5000 OCT has been a ubiquitous screening device for diagnosing posterior segment diseases and constitutes a significant share of retrospective longitudinal data. In view of this, there is significant interest in algorithmically improving the quality of Cirrus OCT images to bring them at par with the quality of EDI-OCT images to improve screening and disease management. Further, improving the quality of cirrus OCT images may enable building accurate disease prediction models leveraging retrospective longitudinal data. Against this backdrop, we propose a novel Cycle Generative Adversarial Networks (Cycle-GANs) based approach to enhance the quality of Cirrus OCT images comparable to that of high-quality EDI-OCT images. In particular, noting the EDI-OCT acquisition process where a high-resolution image is generated by taking the average of scans taken at the scanning location, we attempt a novel approach of mimicking it by taking a novel approach of averaging the model outputs at different epochs of the training, to generate an average synthesized high-quality cirrus image. Further, we have also incorporated structural similarity (SSIM) loss in the loss function to ensure the structural information in the original and enhanced images are preserved. We qualitatively evaluated the performance of the generated enhanced-quality images. In particular, two clinicians have independently graded 45 images for various categories to diagnose retinal diseases and reported a significant improvement in the quality of generated images.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.002 | 0.003 |
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; both teacher heads agree on what is shown here.
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".