Bilateral Papilledema and Partially Empty Sella in a Woman in Her 30s

Abstract

A woman in her 30s was referred for evaluation of bilateral papilledema and progressive visual loss in the right eye, which deteriorated to blindness in 3 months. She denied any other neurologic symptoms, such as headache, nausea, vomiting, seizures, and paresis. She had a 4-year history of Behçet disease (BD), and her condition had been stabilized with thalidomide. Her optic neuropathy of BD had been treated with high-dose intravenous methylprednisolone for 4 days followed by oral prednisone for a month without improvement.Her best-corrected visual acuity was no light perception OD and 20/20 OS. Intraocular pressure was 21 mm Hg OD and 19 mm Hg OS. Fundus examination revealed papilledema in both eyes (Figure 1A). No obvious sign of optic disc ischemia or active uveitis was observed by fundus fluorescein angiography bilaterally. Humphrey visual fields showed complete visual field loss in the right eye and peripheral visual field defects in the left eye. Her anterior segment was normal in both eyes except for relative afferent pupillary defect in the right eye. Cranial nerve examination was otherwise unremarkable. Magnetic resonance imaging of the brain showed no remarkable abnormalities except for partially empty sella (Figure 1B). She had normal blood pressure levels. Results of laboratory testing regarding systemic infection were negative, and her serum D-dimer level was normal. Diagnosis Cerebral venous thrombosis What to Do Next C. Obtain cerebral magnetic resonance venography or computed tomographic venographyClinical Review & Education

Need for Custom Artificial Intelligence Chatbots in Ophthalmology

“…Although the optimized LLM developed by Luo and colleagues outperformed the baseline model Baichuan-13B across 300 clinical ophthalmology questions, it did not surpass the performance of ChatGPT-4. Because the performance of AI models is heavily dependent on the quality of their training data, it is essential for the knowledge corpus of novel optimized models to expand on the set of information available to previous models.…”

“…In order to overcome the current gaps in ophthalmic knowledge within publicly available large language models (LLM), there is a need to develop new optimized AI tools that are tailored for ophthalmic applications. Hence, we commend Luo and colleagues for their initiative of developing a retrieval-augmented LLM framework, leveraging a curated dataset containing over 30 000 data points of ophthalmic knowledge to enhance the accuracy of a baseline generative AI model called Baichuan-13B. The comprehensive ophthalmic dataset used to train the custom LLM presented by Luo and colleagues primarily consisted of Chinese language data, which may limit the model’s applicability to other languages and practice settings.…”

“…Hence, we commend Luo and colleagues for their initiative of developing a retrieval-augmented LLM framework, leveraging a curated dataset containing over 30 000 data points of ophthalmic knowledge to enhance the accuracy of a baseline generative AI model called Baichuan-13B. The comprehensive ophthalmic dataset used to train the custom LLM presented by Luo and colleagues primarily consisted of Chinese language data, which may limit the model’s applicability to other languages and practice settings. At the same time, this work sets an important precedent that we hope may inspire the development of additional specialized LLMs in ophthalmology, particularly tools that can offer nuanced solutions in education, training, clinical decision-making, and research across ophthalmic subspecialties.…”