Determinants of U.S. Health Insurance Charges: Evidence from Multivariate Regression

Abstract

With the rapid development of artificial intelligence, especially generative models and multimodal large models, medical artificial intelligence has gradually moved from the early era of single-modal image recognition and text classification to the era of multimodal modeling. Medical data is inherently multimodal, including various modalities such as images, clinical texts, structured data, and genetic and signal information. For instance, in the diagnosis of lung diseases, multimodal models can integrate chest CT images with patients’ electronic health records (EHR) and medical history texts to quickly locate the lesions and generate preliminary diagnostic suggestions; in the orthopedic treatment scenarios, the models can combine X-ray images with surgical record texts to assist doctors in formulating personalized surgical plans. How to effectively integrate these modalities, and conduct tasks such as diagnostic assistance, report generation, multi-round questioning, pathological explanation and reasoning, has been a focus of research in recent years. This paper systematically reviews the development path of medical multimodal models, summarizes the changes in the capabilities of mainstream methods and the limitations of datasets, and looks forward to the challenges and future trends for practical deployment in clinical settings.

Ähnliche Arbeiten