Researchers at the International Centre for Translational Eye Research (ICTER) have developed an optical coherence tomography (OCT) system that measures the cornea's biomechanical behavior at nine locations simultaneously. By combining ultrafast OCT imaging with a gentle air-puff stimulus, the prototype captures how different regions of the cornea deform in real time.
According to the researchers, the technology could improve the early detection of keratoconus, support treatment monitoring and provide clinicians with more detailed biomechanical information than current diagnostic methods. The study, published in Biomedical Optics Express, illustrates how OCT may evolve from a structural imaging technique into a tool for functional biomechanical assessment.
Looking beyond corneal shape
Current ophthalmic imaging systems routinely assess corneal thickness, curvature and topography. While these measurements are essential for diagnosing eye diseases and planning refractive surgery, they primarily reveal structural changes. In disorders such as keratoconus, however, localized weakening of corneal tissue may occur before visible changes in shape become apparent. Earlier identification of these biomechanical abnormalities could allow clinicians to intervene sooner and potentially slow disease progression.
Existing non-contact biomechanical tests generally use an air puff to deform the cornea while imaging records its response. Although clinically valuable, these systems typically measure the cornea as a whole or within a single imaging plane, making subtle local weaknesses more difficult to detect.
To overcome this limitation, the ICTER researchers developed a system capable of measuring the biomechanical response simultaneously at one central and eight peripheral locations.
Ultrafast imaging
The new technology builds on OCT but introduces a multi-beam optical design. Instead of scanning the cornea point by point, multiple OCT beams acquire data from nine locations simultaneously during a single air-puff event. This innovation addresses a key challenge. Corneal deformation lasts only about 20 milliseconds, meaning even small delays between sequential measurements can reduce accuracy due to eye movements, blinking or tear-film instability.
The prototype records corneal motion every 10 microseconds—equivalent to 100,000 measurements per second across all nine locations. According to the researchers, simultaneous acquisition produces a more faithful representation of corneal biomechanics because every measurement reflects exactly the same mechanical event.
The team also showed why imaging speed matters. When temporal resolution was reduced, errors in estimating both the magnitude and location of biomechanical asymmetry increased substantially. This suggests that ultrafast acquisition is essential for accurately identifying localized tissue weakness.
A new biomarker
Recording deformation at nine locations generates large amounts of data. To make these clinically useful, the researchers introduced a new parameter: the asymmetry vector. This biomarker summarizes two aspects of corneal biomechanics. Its magnitude reflects the degree of mechanical imbalance, while its direction identifies where the greatest deformation occurs.
Rather than describing only the overall flexibility of the cornea, the asymmetry vector pinpoints localized biomechanical weakness. In patients with keratoconus, it consistently pointed toward regions showing thinning, increased curvature and abnormal posterior elevation.
The researchers emphasize that this approach complements rather than replaces existing corneal topography and tomography. Conventional imaging describes anatomy, whereas the OCT system evaluates how the tissue behaves mechanically under stress. Combining both types of information could provide clinicians with a more complete picture of disease development.
Clinical potential
During validation, the researchers observed an unexpected phenomenon: after modifying the air-delivery system, the cornea occasionally displayed two distinct deformation phases following a single air pulse. This previously unreported "dual-indentation response" highlights how simultaneous, high-speed imaging can reveal biomechanical behavior that conventional sequential measurements may miss. Although the biological significance of this finding remains under investigation, it demonstrates the ability of the technology to capture rapid tissue dynamics in unprecedented detail.
The researchers see several potential clinical applications. Earlier detection of keratoconus is perhaps the most promising, since biomechanical changes often precede structural abnormalities. The technology could also support long-term monitoring following corneal cross-linking and improve patient selection for refractive procedures such as LASIK by identifying mechanically vulnerable corneas before surgery.
The current prototype remains a proof of concept and requires further clinical validation before routine use. A larger clinical dataset, including patients with early-stage keratoconus, has already been collected, while the team is developing automated image analysis using machine learning alongside eye-tracking and alignment technologies to streamline future examinations.
Beyond ophthalmology, the researchers believe the underlying principle of simultaneous ultrafast OCT imaging could eventually be applied to other soft tissues where rapid biomechanical events are difficult to measure reliably. If clinical studies confirm its diagnostic value, the technology could become an important addition to the ophthalmic imaging toolbox, enabling earlier diagnosis, more precise disease monitoring and better-informed treatment decisions.
Deep-learning models
In 2024, researchers evaluated deep-learning models for diagnosing infectious keratitis (IK), a corneal infection that affects millions of people worldwide and is a leading cause of preventable corneal blindness, particularly in regions with limited access to specialist eye care. The study analyzed 35 studies involving more than 136,000 corneal images. The AI models matched or slightly outperformed ophthalmologists, achieving a sensitivity of 89.2% and a specificity of 93.2%, compared with 82.2% and 89.6%, respectively, for clinicians.
The models also accurately distinguished between healthy eyes, infected corneas and different causes of keratitis, including bacterial and fungal infections. While the findings highlight AI's potential to support faster and more reliable diagnosis, the researchers emphasize that further external validation and more diverse datasets are needed before the technology can be widely implemented in routine clinical practice.
