THE DEEPER INTO technology we go, the more expansive the opportunities in front of us. Advanced technological tools like artificial intelligence (AI) could help us more effectively diagnose and manage our patients’ vision. Some practitioners are looking into using tools available today, coupled with research findings, to teach AI how to analyze patient data.
A tech breakthrough in the treatment of ocular surface disease is by no means a surprise. A significant problem that clinicians and researchers have is how to agree on what dry eye is, who has dry eye, and how to treat those patients. Dry eye is an incredibly subjective condition. Eyecare providers used to rely solely on a fluorescein or Schirmer strip for dry eye diagnostics, but they now have instrumentation that can give them a more objective understanding of what is happening on the surface of the eye.
The most recent report from the Tear Film & Ocular Surface Society Dry Eye Workshop (TFOS DEWS III) highlights diagnostic testing innovations in its 3-prong approach to diagnosing dry eye (Wolffsohn et al, 2025). In this approach, osmolarity and noninvasive tear breakup time, both objective measures, have more importance. Furthermore, research has demonstrated that machine learning can enhance diagnostic precision.
One study utilized machine learning (ML) modeling to predict tear osmolarity values based on meibomian gland structure and noninvasive tear breakup (Garaszczuk et al, 2024). The model was correct 80% of the time when predicting the correct category of osmolarity.
Another fascinating publication reviewed 47 studies to assess the accuracy of AI algorithms in diagnosing dry eye disease (Heidari et al, 2024). It revealed an approximate 92% accuracy rate of AI models used in previous studies to identify dry eye disease. These AI models analyzed the results of various dry eye clinical tests collectively to make a diagnosis. The review found that studies that included image-based (objective) analysis had high accuracy. While AI is limited by what is already understood in the research world, it is exciting that AI may be able to provide a methodology for clinicians to use to aid in diagnosing disorders in the future.
AI-based electronic health records (EHRs) are currently available that incorporate data based on diagnostic criteria and treatments to organize data sets and assist in treatment decisions. It is possible that future EHR systems could evolve to utilize objective data from patient encounters, apply algorithms from validated research, and inform practitioners’ treatment decisions.
Currently, it is possible for eyecare practitioners to use task-specific conversational AI created through existing AI platforms to build a version of such a system. Using research papers that are uploaded or treatment algorithms that are seen (ie, if corneal staining is present, prescribe an immunomodulator), practitioners can design AI to come up with an instruction and treatment plan based on a specific patient’s clinical data.
While building an AI system for this purpose is akin to a hobbyist playing with programming on a computer in the early days, the process should take minimal time and could give eyecare providers a set of parameters to help guide their decision-making. With time, practitioners will likely be sharing their research-based, task-specific conversational AI systems, and it is possible they will eventually become available for use with EHR software.
Technology is booming. While AI has hardly begun to move, it has great potential to change the way practitioners diagnose and treat patients. Using more objective testing, as highlighted in the new TFOS DEWS III report, will aid clinicians in diagnosis—and may be used in conjunction with AI to provide patients with better outcomes.
References
1.Wolffsohn JS, Benítez-Del-Castillo JM, Loya-García D, et al; TFOS collaborator group. TFOS DEWS III diagnostic methodology. Am J Ophthalmol. 2025:S0002-9394(25)00275-2. doi: 10.1016/j.ajo.2025.05.033
2. Garaxzczuk IK. Romanos-Ibanez M, Consejo A. Machine learningbased prediction of tear osmolarity for contact lens practice. Ophthalmic Physiol Opt. 2024;44(4):727-736. doi: 10.1111/opo.13302
3. Heidari Z, Hashemi H, Sotude D, et al. Applications of artificial intelligence in diagnosis of dry eye disease: a systematic review and meta-analysis. Cornea. 2024;43(10):1310-1318.
