Neu-RadBERT for Enhanced Diagnosis of Brain Injuries and Conditions
Manpreet Singh (1), Sean Macrae (2), Pierre-Marc Williams (2), Nicole Hung (2), Sabrina Araujo de Franca (1), Laurent Letourneau-Guillon (2, 3), François-Martin Carrier (2, 4), Bang Liu (5), Yiorgos Alexandros Cavayas (1, 2, 6) ((1) Équipe de Recherche en Soins Intensifs, Centre de recherche du Centre intégré universitaire de santé et de services sociaux du Nord-de-l'Île-de-Montréal (2) Faculté de Médecine, Université de Montréal (3) Department of Radiology, Centre Hospitalier de l'Université de Montréal (4) Department of Anesthesia, Centre Hospitalier de l'Université de Montréal (5) Applied Research in Computer Linguistics Laboratory, Department of Computer Science and Operations Research, Université de Montréal (6) Division of Critical Care Medicine, Department of Medicine, Hôpital du Sacré-Cœur de Montréal)
目标:我们试图开发一种分类算法,从接受侵入性机械通气的急性呼吸衰竭(ARF)患者进行脑成像的自由文本放射学报告中提取诊断。 方法:我们开发和微调了基于BERT的模型Neu-RadBERT,对非结构化放射学报告进行分类。 我们从MIMIC-IV数据库中提取了所有脑成像报告(计算机断层扫描和磁共振成像),在ARF患者中执行。 初始手动标签是在针对各种大脑异常的报告子集上进行的,然后使用三种策略对Neu-RadBBERT进行微调:1)基线RadBERT,2)带有蒙面语言建模(MLM)预训练的Neu-RadBERT,以及3)Neu-RadBBERT与MLM预训练和过度采样以解决数据偏斜。 我们将该模型的性能与Llama-2-13B(一种自动回归LLM)进行了比较。 结果:Neu-RadBERT模型,特别是过度采样,与基线RadBERT相比,诊断准确性显着提高,达到98.0
Objective: We sought to develop a classification algorithm to extract diagnoses from free-text radiology reports of brain imaging performed in patients with acute respiratory failure (ARF) undergoing invasive mechanical ventilation. Methods: We developed and fine-tuned Neu-RadBERT, a BERT-based model, to classify unstructured radiology reports. We extracted all the brain imaging reports (computed tomography and magnetic resonance imaging) from MIMIC-IV database, performed in patients with ARF. I...
