CNS-Obsidian: A Neurosurgical Vision-Language Model Built From Scientific Publications

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Alyakin, Anton, Stryker, Jaden, Alber, Daniel Alexander, Lee, Jin Vivian, Sangwon, Karl L., Duderstadt, Brandon, Save, Akshay, Kurland, David, Frome, Spencer, Singh, Shrutika, Zhang, Jeff, Yang, Eunice, Park, Ki Yun, Orillac, Cordelia, Valliani, Aly A., Neifert, Sean, Liu, Albert, Patel, Aneek, Livia, Christopher, Lau, Darryl, Laufer, Ilya, Rozman, Peter A., Hidalgo, Eveline Teresa, Riina, Howard, Feng, Rui, Hollon, Todd, Aphinyanaphongs, Yindalon, Golfinos, John G., Snyder, Laura, Leuthardt, Eric, Kondziolka, Douglas, Oermann, Eric Karl
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918216756887552
author Alyakin, Anton
Stryker, Jaden
Alber, Daniel Alexander
Lee, Jin Vivian
Sangwon, Karl L.
Duderstadt, Brandon
Save, Akshay
Kurland, David
Frome, Spencer
Singh, Shrutika
Zhang, Jeff
Yang, Eunice
Park, Ki Yun
Orillac, Cordelia
Valliani, Aly A.
Neifert, Sean
Liu, Albert
Patel, Aneek
Livia, Christopher
Lau, Darryl
Laufer, Ilya
Rozman, Peter A.
Hidalgo, Eveline Teresa
Riina, Howard
Feng, Rui
Hollon, Todd
Aphinyanaphongs, Yindalon
Golfinos, John G.
Snyder, Laura
Leuthardt, Eric
Kondziolka, Douglas
Oermann, Eric Karl
author_facet Alyakin, Anton
Stryker, Jaden
Alber, Daniel Alexander
Lee, Jin Vivian
Sangwon, Karl L.
Duderstadt, Brandon
Save, Akshay
Kurland, David
Frome, Spencer
Singh, Shrutika
Zhang, Jeff
Yang, Eunice
Park, Ki Yun
Orillac, Cordelia
Valliani, Aly A.
Neifert, Sean
Liu, Albert
Patel, Aneek
Livia, Christopher
Lau, Darryl
Laufer, Ilya
Rozman, Peter A.
Hidalgo, Eveline Teresa
Riina, Howard
Feng, Rui
Hollon, Todd
Aphinyanaphongs, Yindalon
Golfinos, John G.
Snyder, Laura
Leuthardt, Eric
Kondziolka, Douglas
Oermann, Eric Karl
contents General-purpose VLMs demonstrate impressive capabilities, but their opaque training on uncurated internet data poses critical limitations for high-stakes decision-making, such as in neurosurgery. We present CNS-Obsidian, a neurosurgical VLM trained on peer-reviewed literature, and demonstrate its clinical utility versus GPT-4o in a real-world setting. We compiled 23,984 articles from Neurosurgery Publications journals, yielding 78,853 figures and captions. Using GPT-4o and Claude Sonnet-3.5, we converted these into 263,064 training samples across three formats: instruction fine-tuning, multiple-choice questions, and differential diagnosis. We trained CNS-Obsidian, a fine-tune of the 34-billion parameter LLaVA-Next model. In a blinded, randomized trial at NYU Langone Health (Aug 30-Nov 30, 2024), neurosurgery consultations were assigned to either CNS-Obsidian or a HIPAA-compliant GPT-4o endpoint as diagnostic co-pilot after consultations. Primary outcomes were diagnostic helpfulness and accuracy, assessed via user ratings and presence of correct diagnosis within the VLM-provided differential. CNS-Obsidian matched GPT-4o on synthetic questions (76.13% vs 77.54%, p=0.235), but only achieved 46.81% accuracy on human-generated questions versus GPT-4o's 65.70% (p<10-15). In the randomized trial, 70 consultations were evaluated (32 CNS-Obsidian, 38 GPT-4o) from 959 total consults (7.3% utilization). CNS-Obsidian received positive ratings in 40.62% of cases versus 57.89% for GPT-4o (p=0.230). Both models included correct diagnosis in approximately 60% of cases (59.38% vs 65.79%, p=0.626). Domain-specific VLMs trained on curated scientific literature can approach frontier model performance despite being orders of magnitude smaller and less expensive to train. This establishes a transparent framework for scientific communities to build specialized AI models.
format Preprint
id arxiv_https___arxiv_org_abs_2502_19546
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CNS-Obsidian: A Neurosurgical Vision-Language Model Built From Scientific Publications
Alyakin, Anton
Stryker, Jaden
Alber, Daniel Alexander
Lee, Jin Vivian
Sangwon, Karl L.
Duderstadt, Brandon
Save, Akshay
Kurland, David
Frome, Spencer
Singh, Shrutika
Zhang, Jeff
Yang, Eunice
Park, Ki Yun
Orillac, Cordelia
Valliani, Aly A.
Neifert, Sean
Liu, Albert
Patel, Aneek
Livia, Christopher
Lau, Darryl
Laufer, Ilya
Rozman, Peter A.
Hidalgo, Eveline Teresa
Riina, Howard
Feng, Rui
Hollon, Todd
Aphinyanaphongs, Yindalon
Golfinos, John G.
Snyder, Laura
Leuthardt, Eric
Kondziolka, Douglas
Oermann, Eric Karl
Artificial Intelligence
Computation and Language
Human-Computer Interaction
General-purpose VLMs demonstrate impressive capabilities, but their opaque training on uncurated internet data poses critical limitations for high-stakes decision-making, such as in neurosurgery. We present CNS-Obsidian, a neurosurgical VLM trained on peer-reviewed literature, and demonstrate its clinical utility versus GPT-4o in a real-world setting. We compiled 23,984 articles from Neurosurgery Publications journals, yielding 78,853 figures and captions. Using GPT-4o and Claude Sonnet-3.5, we converted these into 263,064 training samples across three formats: instruction fine-tuning, multiple-choice questions, and differential diagnosis. We trained CNS-Obsidian, a fine-tune of the 34-billion parameter LLaVA-Next model. In a blinded, randomized trial at NYU Langone Health (Aug 30-Nov 30, 2024), neurosurgery consultations were assigned to either CNS-Obsidian or a HIPAA-compliant GPT-4o endpoint as diagnostic co-pilot after consultations. Primary outcomes were diagnostic helpfulness and accuracy, assessed via user ratings and presence of correct diagnosis within the VLM-provided differential. CNS-Obsidian matched GPT-4o on synthetic questions (76.13% vs 77.54%, p=0.235), but only achieved 46.81% accuracy on human-generated questions versus GPT-4o's 65.70% (p<10-15). In the randomized trial, 70 consultations were evaluated (32 CNS-Obsidian, 38 GPT-4o) from 959 total consults (7.3% utilization). CNS-Obsidian received positive ratings in 40.62% of cases versus 57.89% for GPT-4o (p=0.230). Both models included correct diagnosis in approximately 60% of cases (59.38% vs 65.79%, p=0.626). Domain-specific VLMs trained on curated scientific literature can approach frontier model performance despite being orders of magnitude smaller and less expensive to train. This establishes a transparent framework for scientific communities to build specialized AI models.
title CNS-Obsidian: A Neurosurgical Vision-Language Model Built From Scientific Publications
topic Artificial Intelligence
Computation and Language
Human-Computer Interaction
url https://arxiv.org/abs/2502.19546