Physics-aware learning and domain-specific loss design in ophthalmology

• Published in: Medical image analysis 2022-02, Vol.76, p.102314-102314, Article 102314
• Main Authors: Burwinkel, Hendrik, Matz, Holger, Saur, Stefan, Hauger, Christoph, Trost, Michael, Hirnschall, Nino, Findl, Oliver, Navab, Nassir, Ahmadi, Seyed-Ahmad
• Format: Article
• Language: English
• Subjects: Biometrics; Blindness; Cataracts; Datasets; Domain prior incorporation; Eye; Eye lens; Intraocular lenses; IOL calculation; Machine learning; Ophthalmology; Optics; Physics-based learning; Surgery; Transfer learning
• ISSN: 1361-8415; 1361-8423
• DOI: 10.1016/j.media.2021.102314

•Current intraocular lens calculation uses available knowledge sub-optimally.•Proposed method OpticNet combines physical modelling and data-driven ML.•Novel physics-based unsupervised loss function drives training process.•Neural network weights internalize optical raytracing physics of the eyeball.•Fine-tuning on small amounts of real surgical data outperforms state of the art. [Display omitted] The human cataract, a developing opacification of the human eye lens, currently constitutes the world’s most frequent cause for blindness. As a result, cataract surgery has become the most frequently performed ophthalmic surgery in the world. By removing the human lens and replacing it with an artificial intraocular lens (IOL), the optical system of the eye is restored. In order to receive a good refractive result, the IOL specifications, especially the refractive power, have to be determined precisely prior to surgery. In the last years, there has been a body of work to perform this prediction by using biometric information extracted from OCT imaging data, recently also by machine learning (ML) methods. Approaches so far consider only biometric information or physical modelling, but provide no effective combination, while often also neglecting IOL geometry. Additionally, ML on small data sets without sufficient domain coverage can be challenging. To solve these issues, we propose OpticNet, a novel optical refraction network based on an unsupervised, domain-specific loss function that explicitly incorporates physical information into the network. By providing a precise and differentiable light propagation eye model, physical gradients following the eye optics are backpropagated into the network. We further propose a new transfer learning procedure, which allows the unsupervised pre-training on the optical model and fine-tuning of the network on small amounts of surgical patient data. We show that our method outperforms the current state of the art on five OCT-image based data sets, provides better domain coverage within its predictions, and achieves better physical consistency.