Improvement and validation of high precision ocular oximetry using a convolutional neural network algorithm and a phantom eye
Bibliographic record
Abstract
Abstract Retinal oximetry is a non‐invasive imaging technology that enables the measurement of oxygen saturation (StO 2 ) in the eye fundus. The goal of this research was to validate a convolutional neural network (CNN) algorithm designed to calculate and improve the precision of StO 2 measurements from diffuse reflectance spectra (DRS) taken on the optic nerve head (ONH). Zilia’s multi‐wavelength retinal oximetry device was used to acquire experimental spectra from the ONH which allowed us to simulate reasonable digital spectra replicates with known absorber concentrations. These spectra were used to train a novel machine learning algorithm based on CNN. The device was then used to acquire diffuse reflectance spectra on several ONH‐mimicking liquid optical phantoms (phantom eye) with dynamic oxygenation cycles between 0% ‐ 100% in order to validate the improvements of this CNN on experimental data. Measurements were made simultaneously with gold standard devices for comparison. The procedure was then repeated with several cataract‐simulating contact lenses integrated in the optical path to show the robustness of oximetry measurements. We found good agreement in StO 2 measurements between the results obtained with the Zilia device using the CNN algorithm and the gold standard references in all phantom cases. Applying the various algorithms to data acquired on the validation phantom show the marked improvements in using the CNN on experimental data, validating its high potential for clinical use. We specifically show strong robustness in precision, even when cataract‐simulating lenses were used. We present further validation that the oximetry device and CNN algorithm produces reliable, precise measurements even under conditions where blood volume fractions vary, optical scattering changes, and cataract‐simulating contact lenses are included.
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 | 0.000 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, 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".