Reduced order modelling of air puff test for corneal material characterisation

Osama M. Maklad, Muting Hao · 2025

Read Reduced order modelling of air puff test for corneal material characterisation by Osama M. Maklad, Muting Hao (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-22495.txt.

Introduction

Osama Maklad and Muting Hao speed up computer modeling of the air puff test used to probe the cornea, in a 2025 paper. The test fires a jet of air at the eye and watches the clear front window bend, which reveals its stiffness and the pressure inside. Full simulations that couple air and tissue take about two days to run.

Their reduced model trims that to a twelve-minute tissue-only run, and a physics-informed neural network is added to stretch it across larger data sets. A second network learns how pressure and bending spread over the surface in space and time. The authors aim at faster, sharper estimates of eye pressure and corneal traits for clinic use.

Catalog record

CREATOROsama M. Maklad, Muting Hao
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-22495

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Reduced order modelling of air puff test for corneal material characterisation Osama M. Maklad, Muting Hao Abstract Models of the fluid-structure interaction (FSI) model for the air puff test were analysed. Using Abaqus, the air puff test is applied to eyes with varying biomechanical parameters, such as material properties, corneal thickness, and radius. A reduced order model of the air puff (a turbulent impinging jet) has been acquired to decrease simulation time from 48 hours for the FSI model to approximately 12 minutes for the finite element analysis (FEA) model alone. To further accelerate simulations and improve model accuracy, Physics-Informed Neural Networks (PINNs) will be integrated with the reduced-order model. This hybrid approach will help expand the model to a larger dataset, enhancing intraocular pressure (IOP) estimation accuracy and the corneal material properties algorithm through inverse FEA. Additionally, a neural network (NN) framework with embedded Gaussian-modulated waveforms is proposed to model the pressure and deformation distributions on the corneal surface as functions of spatial and temporal parameters. By learning the relationship between corneal biomechanical inputs such as Corneal Central Thickness (CCT), Intraocular Pressure (IOP), and baseline properties ( μ \mu ), and the governing coefficients of pressure and deformation, the network accurately reconstructs the result that matches well with the high-fidelity CFD data. This approach can quickly capture the distribution of pressure and deformation. It can also provide insights into the distinct spatial and temporal dynamics of pressure and deformation, giving a more comprehensive understanding of fluid-structure interaction phenomena in the air puff test. Abstract Keywords: Air puff test, Intraocular Pressure (IOP), Ocular biomechanics, Fluid-Structure Interaction (FSI), Reduced order modelling, Machine Learning (ML), Gradient Boosting Regressor (GBR) Nomenclature IOP = Intraocular pressure CCT = Central corneal thickness \mu = Corneal stiffness parameter GBR = Gradient boosting regression 1 Introduction In ophthalmology, the non-contact air puff tonometry test is widely used to measure the biomechanical properties of the human cornea and intraocular pressure (IOP). Accurate IOP measurement is critical in evaluating patients at risk of eye diseases such as glaucoma, where elevated IOP can lead to optic nerve damage. Glaucoma is one of the leading causes of asymptomatic permanent blindness in the developed world. A common cause of IOP increase is the improper drainage of aqueous humour due to trabecular meshwork blockage [ 1 ] . The Ocular Response Analyzer (ORA) was the first tonometry device that utilized an air puff to assess ocular biomechanical properties, using a dynamic infrared signal analysis [ 2 ] . Later, the Corvis-ST tonometer was developed, incorporating an ultra-high-speed Scheimpflug camera for better visualization of corneal deformation [ 3 ] . The Corvis-ST applies a concentrated air puff to the centre of the cornea, deforming its geometry, which subsequently regains its original shape due to its elasticity. Using image processing, corneal deformation is recorded to estimate biomechanical properties and IOP through a programmed parametric equation. Keratoconus, a degenerative eye disease, is characterized by the progressive thinning of the cornea, resulting in a cone-shaped protrusion [ 4 ] . This condition significantly alters the cornea’s thickness, shape, and biomechanical properties [ 5 ] , often causing irregular astigmatism and blurry vision. In patients with keratoconus, tonometry measurements of IOP tend to be lower due to the strong correlation between IOP readings and the altered biomechanical properties, particularly corneal central thickness (CCT) [ 6 , 7 ] . Understanding these biomechanical changes is crucial for clarifying the pathophysiology and aetiology of keratoconus, aiding in its treatment [ 8 ] . Accurate in vivo measurements of both IOP and corneal biomechanical parameters are essential. However, the challenge lies in the mutual dependence between these parameters, making it difficult to isolate the effects of IOP from biomechanical properties like thickness and material stiffness on corneal response [ 3 ] . Solving an inverse problem offers a potential solution to this challenge, allowing for a more accurate assessment of corneal material behaviour based on improved IOP values [ 9 , 10 ] . The study of fluid-structure interaction (FSI) between the air puff and the cornea is necessary to address the correlation between IOP and corneal parameters and improve IOP estimation accuracy [ 11 ] . To achieve this, the air puff pressure distribution profiles on the cornea must be considered when measuring corneal deformation [ 3 ] . The pressure exerted on the cornea dynamically changes in response to corneal deformation, which can alter clinical interpretations [ 12 ] . This effect termed the Corneal Load Alteration with Surface Shape (CLASS), highlights the importance of obtaining patient-specific air puff pressure distributions [ 12 , 3 , 9 ] . While the co-simulation of FSI has provided accurate values for pressure distribution and corneal deformation [ 9 , 13 , 14 ] , it remains time-consuming. The evolution of FSI approaches has led to precise estimations of air jet pressure changes with corneal characteristics, resulting in biomechanically corrected equations for accurate IOP and corneal material stress-strain indices [ 3 , 9 , 12 ] . However, predicting air puff pressure in the computational fluid dynamics (CFD) model of FSI is computationally expensive, taking considerable time for patient-specific geometries. The primary goal of this study is to investigate how individual corneal characteristics affect air puff pressure measurements and develop a machine learning (ML) algorithm to reduce computational costs. While numerical models of fluid flow are central to research on physical and mechanical phenomena, especially those involving FSI [ 9 , 12 , 13 ] , recent studies [ 15 , 16 , 17 ] suggest that ML can replace time-consuming numerical solvers, providing reduced-order models, improved optimization, and lower computational costs. Therefore, we have developed a supervised regression ML algorithm using the Gradient Boosting Regressor (GBR) to estimate time-dependent air puff corneal pressure distribution profiles, replacing the CFD model in the FSI co-simulation of [ 9 ] to reduce computational costs. This study focuses on the effect of changing corneal parameters (IOP, CCT, and material stiffness) on air puff pressure profiles, proposing a GBR-based algorithm for estimating patient-specific air puff pressure loads, and applying the predicted loads to the finite element model of the eye in ABAQUS software to simulate corneal deformation without the need for a time-consuming CFD model. To further improve the efficiency and accuracy of corneal response predictions, this study introduces a hybrid Neural Network (NN)with parametric Embedding framework to model dynamic pressure and deformation distributions on the corneal surface. Observed patterns of pressure and deformation resemble Gaussian-modulated waveforms [ 18 ] , motivating the use of a parametric model characterized by five coefficients: amplitude ( A ), spatial scaling ( \beta ), temporal characteristics ( \mu and \sigma ), and spatial attenuation ( \alpha ). In this study, the NN is trained to map biomechanical input parameters—Corneal Central Thickness (CCT), Intraocular Pressure (IOP), and baseline material properties ( \mu_{0} )—to these coefficients, enabling efficient reconstruction of pressure and deformation distributions. This approach accelerates computational workflows, replacing the need for expensive CFD simulations. It can also provide new insights into the spatial and temporal dynamics of corneal responses. 2 Method This section outlines the dataset used in our approach and describes the algorithm applied for learning and prediction.

Read the complete text