Recent advancements in Artificial Intelligence (AI) have profoundly
influenced medical fields, by providing tools to reduce clinical workloads.
However, most AI models are constrained to execute uni-modal tasks, in stark
contrast to the comprehensive approaches utilized by medical professionals. To
address this, here we present RO-LLaMA, a versatile generalist large language
model (LLM) tailored for the field of radiation oncology. This model seamlessly
covers a wide range of the workflow of radiation oncologists, adept at various
tasks such as clinical report summarization, radiation therapy plan suggestion,
and plan-guided therapy target volume segmentation. In particular, to maximize
the end-to-end performance, we further present a novel Consistency Embedding
Fine-Tuning (CEFTune) technique, which boosts LLM's robustness to additional
errors at the intermediates while preserving the capability of handling clean
inputs, and creatively transform this concept into LLM-driven segmentation
framework as Consistency Embedding Segmentation (CESEG). Experimental results
on multi-centre cohort sets demonstrate our proposed RO-LLaMA's promising
performance for diverse tasks with generalization capabilities.

近期人工智能在医疗领域有了显著进展，本研究提出 RO-LLaMA，一个通用的大型语言模型，特为放射肿瘤学领域量身定制，能够在临床报告总结、放疗方案建议等任务中表现出卓越的性能，并且通过 CEFTune 技术进一步提高了模型的鲁棒性，同时保持了处理干净输入的能力，还将此概念创造性地转化为 LLM 驱动的分割框架 CESEG。实验结果表明，RO-LLaMA 在多中心队列数据集上展现了出色的性能和泛化能力。

RO-LLaMA：通过噪声增强和一致性正则化实现放射治疗的综合型医学语言模型

RO-LLaMA: Generalist LLM for Radiation Oncology via Noise Augmentation  and Consistency Regularization

Decision support systems play an important role in medical fields as they can
augment clinicians to deal more efficiently and effectively with complex
decision-making processes. In the diagnosis of headache disorders, however,
existing approaches and tools are still not optimal. On the one hand, to
support the diagnosis of this complex and vast spectrum of disorders, the
International Headache Society released in 1988 the International
Classification of Headache Disorders (ICHD), now in its 3rd edition: a 200
pages document classifying more than 300 different kinds of headaches, where
each is identified via a collection of specific nontrivial diagnostic criteria.
On the other hand, the high number of headache disorders and their complex
criteria make the medical history process inaccurate and not exhaustive both
for clinicians and existing automatic tools. To fill this gap, we present
HEAD-ASP, a novel decision support system for the diagnosis of headache
disorders. Through a REST Web Service, HEAD-ASP implements a dynamic
questionnaire that complies with ICHD-3 by exploiting two logical modules to
reach a complete diagnosis while trying to minimize the total number of
questions being posed to patients. Finally, HEAD-ASP is freely available
on-line and it is receiving very positive feedback from the group of
neurologists that is testing it.

本文提出了 HEAD-ASP，一个新型的支持头痛诊断的决策辅助系统，通过一个基于 REST 的 Web 服务，利用两个逻辑模块实现了动态问卷，以尽可能少的提问问题，遵循 ICHD-3 分类诊断，已在神经学家小组中得到了很好的反馈。