KEDOP: Keratoconus early detection of progression using tomography images
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
PURPOSE: To investigate a method to identification of early progression of keratoconus using deep learning neural networks. METHODS: Retrospective evaluation of medical records of patients with progressive keratoconus and had more than one followup visits. Images extracted from the single scheimplug analyzer for analysis were captured during the patient visits. The baseline progression of keratoconus is detected by a change in flat or steep K of ≥1.0D which is labeled as keratometric progression (KP) and progression detected by image based deep learning convolutional neural network (CNN) models, is labeled as latent progression (LP). Patient data consisted of model data (385 eyes of 351patients) to train and test the learning models and prediction data (1331 eyes of 828 patients) to determine the LP based on the learning models. RESULTS: The LP prediction model was able to identify progression at a mean of 11.1 months earlier than KP (p < 0.001). LP prediction model was able to identify progression earlier than KP irrespective of age category, gender, the severity of keratoconus, presenting visual acuity, astigmatism, and spherical equivalent (P < 0.001). When compared to the first visit the corrected distance visual acuity was more stable in 71% of the eyes at LP prediction visit compared to 50% at KP visit (p < 0.001). CONCLUSION: Through this study, we propose a possible solution to address the shortcomings noted in the current approaches of detecting progression relying only on KP. Avoiding bias towards feature selection from tomography images as done in the current study aids in identifying very subtle changes on the images between visits.
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.
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 | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 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 it