Clinical assessment of ex vivo biomechanical properties of the human cataractous lens using machine learning
- 1Department of Biomedical Engineering, School of Life Science Engineering, University of Tehran
- 2Department of Mathematics and Computer Science, Faculty of Sciences, University of Zanjan
- 3Department of Ophthalmology, Imam Hassan University Hospital, Alborz University of Medical Sciences
Abstract
Background Lens stiffness plays a central role in cataract surgery planning, especially for phacoemulsification. Current assessments rely on intraoperative judgment, which limits personalized surgical strategies. This pilot study aimed to develop the first machine learning model using extracapsular cataract extraction–derived ground-truth data to predict the stiffness of cataractous human lenses using routinely available demographic and clinical features. Methods Fifty human lens nuclei were obtained following extracapsular cataract extraction. Predictive features included age, gender, cataract grade, lens diameter, diabetes status, and aspirin use. Each lens was exposed to three compressive loads of 0.2 N, and deformation was measured using ultrasound imaging. Young’s modulus was calculated for each trial. Data preprocessing, exploratory analysis, and model training were performed. Several supervised regression algorithms were developed and compared in terms of predictive accuracy and generalization. Results Among the implemented 10 regression models, the Ridge model demonstrated the highest predictive accuracy. Feature importance analysis identified cataract grade, lens diameter, age, and diabetes status as the strongest predictors of stiffness. Aspirin usage and gender showed relatively weaker contributions to model performance. Conclusions Supervised machine learning appears feasible for estimating lens stiffness from routine demographic and clinical data. Moreover, given the small, nonrepresentative ECCE-based sample, these findings should be regarded as proof-of-concept and exploratory. Validation in substantially larger, independent phacoemulsification-based cohorts is required before any clinical application can be considered.
Related
- Corvis ST Assessment of Corneal Biomechanical Properties in Patients with Type 2 Diabetes Mellitus
- Content Validity and Psychometric Properties Assessment of the Virtual Reality Functional Capacity Assessment Tool and Application in Multiregional Clinical Trials of Cognitive Impairment Associated with Schizophrenia.
- Quantitative MRI Characterization of Fluid-Related Phenotypes in Pediatric Epilepsy: Deep Learning-Assisted Assessment of Enlarged Perivascular Space Burden, Regional Distribution, and Clinical Associations
- Deep-Learning-Based Assessment of Breast Arterial Calcification on Mammography: Technical Validation, Cardiovascular Prognostic Value, and Clinical Translation—A Narrative Review
- The impact of osteoporosis on spinal biomechanical stability in adolescent idiopathic scoliosis: a narrative review of finite element analysis and bone mineral density assessment.
BibTeX
@article{tabesh2026clinical,
title = {{Clinical assessment of ex vivo biomechanical properties of the human cataractous lens using machine learning}},
author = {Hadi Tabesh and Rosa Kheirkhah and Sara Rahmati and Pouya Jozesoleimani and Farrokh Farzad},
journal = {European Journal of Medical Research},
year = {2026},
month = jul,
doi = {10.1186/s40001-026-04946-0},
url = {https://doi.org/10.1186/s40001-026-04946-0},
note = {CC BY},
}
RIS
TY - JOUR
TI - Clinical assessment of ex vivo biomechanical properties of the human cataractous lens using machine learning
AU - Hadi Tabesh
AU - Rosa Kheirkhah
AU - Sara Rahmati
AU - Pouya Jozesoleimani
AU - Farrokh Farzad
PY - 2026
DA - 2026/07/01
JO - European Journal of Medical Research
DO - 10.1186/s40001-026-04946-0
UR - https://doi.org/10.1186/s40001-026-04946-0
AB - Background Lens stiffness plays a central role in cataract surgery planning, especially for phacoemulsification. Current assessments rely on intraoperative judgment, which limits personalized surgical strategies. This pilot study aimed to develop the first machine learning model using extracapsular cataract extraction–derived ground-truth data to predict the stiffness of cataractous human lenses using routinely available demographic and clinical features. Methods Fifty human lens nuclei were obtained following extracapsular cataract extraction. Predictive features included age, gender, cataract grade, lens diameter, diabetes status, and aspirin use. Each lens was exposed to three compressive loads of 0.2 N, and deformation was measured using ultrasound imaging. Young’s modulus was calculated for each trial. Data preprocessing, exploratory analysis, and model training were performed. Several supervised regression algorithms were developed and compared in terms of predictive accuracy and generalization. Results Among the implemented 10 regression models, the Ridge model demonstrated the highest predictive accuracy. Feature importance analysis identified cataract grade, lens diameter, age, and diabetes status as the strongest predictors of stiffness. Aspirin usage and gender showed relatively weaker contributions to model performance. Conclusions Supervised machine learning appears feasible for estimating lens stiffness from routine demographic and clinical data. Moreover, given the small, nonrepresentative ECCE-based sample, these findings should be regarded as proof-of-concept and exploratory. Validation in substantially larger, independent phacoemulsification-based cohorts is required before any clinical application can be considered.
SN - 2047-783X
ER -