{"ok":true,"nota":{"id":"7445a71ea01d18bf","slug":"una-ia-que-analiza-fotos-de-comidas-se-equivoca-bastante-al-estimar-ca-7445a7","titulo":"Una IA que analiza fotos de comidas se equivoca bastante al estimar calorías y nutrientes","resumen":"Evaluaron ocho sistemas de inteligencia artificial multimodal con fotos reales de comidas para estimar calorías, carbohidratos, proteínas y calidad de la dieta. Compararon su precisión y sus limitaciones para personas con diabetes.","fuente":"medrxiv","fuenteNombre":"medRxiv · preprints","color":"#1f7a4d","autor":"Gravellier, L., Zhang, C., Wu, Y., Parameswaran, V., Scheideman, A., Shao, M., Zhang, E., Perelman, D., Gardner, C. D., ","url":"https://www.medrxiv.org/content/10.64898/2026.10.21.26365949v1?rss=1","imagen":"/hoy/portada/7445a71ea01d18bf","tema":"salud","subtemas":["diabetes","nutrición","inteligencia artificial","alimentación","autocontrol de salud"],"entidades":["SeeFood","SNAPMe","medRxiv"],"formato":"dato","entretenimiento":false,"pubTs":"2026-10-24T00:00:00.000Z","indexTs":"2026-10-24T16:44:51.482Z","hash":"6eea10d8dc65cfbf886f95a7b7ef5af9b48fa1a56b4989a482cde8801b740a3a","tsa":{"time":"Oct 24 16:45:15 2026 GMT","provider":"freetsa.org","sobre":"sello"},"stats":{"views":0,"likes":0,"clicks":0,"dwell":0},"tipo":"normal","registro":"hallazgo","porQueImporta":"Cada vez más gente usa una foto del plato para llevar la cuenta de lo que come, algo especialmente útil para quien tiene diabetes y necesita controlar carbohidratos. El estudio muestra que estas herramientas todavía tienen errores importantes al medir lo que hay en el plato, así que no conviene tomar sus cifras como exactas para ajustar la medicación. Para quien usa una app de este tipo en Argentina, la foto puede servir como orientación, pero no reemplaza la consulta con un nutricionista o el médico tratante.","quienFirma":"medRxiv · preprints","confianza":"preprint","consulta":null},"sello":{"hash":"6eea10d8dc65cfbf886f95a7b7ef5af9b48fa1a56b4989a482cde8801b740a3a","sello":"dfb119e85c412097fd02dd3a097227fe0511aeae19297e8dba8d1af4c438739d","beacon":{"drand":{"cadena":"8990e7a9aaed2ffed73dbd7092123d6f289930540d7651336225dc172e51b2ce","ronda":6580975,"aleatorio":"7cce22f6f8f0e18a1e57d3c0d55b5004bf079b568a9a9f143dc9251a504825bd","tiempo":"2026-10-24T16:44:30.000Z"},"nist":{"cadena":2,"pulso":1997681,"salida":"C5248B3EE24B422962809C61551C0ED22A554FF83DF586AF6189AC5F408A5DD6B8269834835CE193FDEC10C180159B7695EE635C161F0BDB8944F4CE7E0AE962","tiempo":"2026-10-24T16:44:00.000Z"},"noAntes":"2026-10-24T16:44:30.000Z"},"verificacion":{"drand":true,"nist":true},"tsa":{"time":"Oct 24 16:45:15 2026 GMT","provider":"freetsa.org","sobre":"sello"},"original":{"titulo":"Image-Based Dietary Assessment Using Multimodal Large Language Models: Performance, Limitations, and Clinical Applicability in Diabetes","descripcion":"Background Image-based dietary assessment is useful for precision diabetes care. Multimodal Large Language Models (mLLMs) may greatly facilitate dietary self-monitoring. However, comprehensive evaluation of their real-world accuracy remains limited. Methods We developed SeeFood, an open-source benchmark, and evaluated eight commercial and open-source mLLMs on 80 real-world meal photographs from the SNAPMe dataset for nutrient and energy estimation (calories, carbohydrates, protein, fat, fiber), food-item identification, and day-level diet quality (Healthy Eating Index, HEI-2020). Three finalists were evaluated on the complete dataset (1680 photographs, 275 participant-days), and the best mod","pubDate":"2026-10-24","url":"https://www.medrxiv.org/content/10.64898/2026.10.21.26365949v1?rss=1"},"indexado":"2026-10-24T16:44:51.482Z"}}