Answers are grounded in this paper and cite their sources.
Relevance. The relevance of the study is due to the large number of studies being conducted — optical coherence tomography of the macular area of the retina, the lack of uniform rules for describing the results obtained during the study, which complicates clinical, organizational and scientific work when using accumulated and newly obtained data. Purpose. To test the hypothesis about the possibility of creating a reference book for describing the results of optical coherence tomography, formulating a standard for describing an optical coherence tomography image using natural language processing (unstructured data previously accumulated in medical information systems). Material and methods. An analysis of 30,000 unstructured text descriptions of medical images of the norm and pathology of the macular of the human retina was carried out. The BERTopic topic modeling technique was applied to the resulting data array. Results. After modeling and initial cleaning, 123 topics were discovered, to which the hierarchical clustering method was applied to combine them into syndrome complexes, indicating the absence or presence of a pathological process with the name of this syndrome complex based on expert assessment. The data obtained is systematized in a table for use in practical activities. A complete directory of sections, signs, syndrome complexes is available for download, use, and integration into medical information systems at the link https://disk.yandex.ru/i/agKdkWGvYyMfaQ. Conclusion. The resulting guide can be useful in practical work: creating medical information systems, medical decision support systems based on neural systems, other deep machine learning methods, the hypothesis about the possibility of constructing reference books by processing large arrays of “natural” language has been confirmed.
No takes yet. Share an insight, caveat, or question.
Federation et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: