MDwAIstScheduler：将设备端语音记录引入临床实践
MDwAIstScheduler: Bringing On-Device Voice Documentation into Clinical Practice
AI总结：
该研究提出设备端临床语音记录工具MDwAIstScheduler，通过设备端运行转录和意图提取，将诊疗语音转化为EHR可审核草稿，减轻医生记录负担并提升诊疗专注度。
AI中文摘要：
临床记录迫使医生在患者和键盘之间分散注意力，大量记录工作变成无报酬的下班后工作。我们提出MDwAIstScheduler，一种低成本的腰带式流程，让医生在诊疗过程中自然说话，生成的药物、过敏、检查/医嘱/转诊、随访安排、生命体征和问题会以可审核的草稿形式进入电子健康记录（EHR）。基于我们早期依赖云语音识别和云语言模型的原型，当前流程完全在设备端运行转录和意图提取。使用医疗领域自动语音识别（ASR）模型和我们为临床动作提取微调的17亿参数语言模型，患者音频或文本不会离开设备，结构化草稿直接写入Elation EHR供医生确认。该工具消除了诊疗过程中的键盘工作，同时不使临床医生脱离记录，让他们专注于最重要的患者护理，同时减轻负担。
英文摘要：
Clinical documentation forces physicians to split attention between the patient and their keyboard, and much of it spills into uncom- pensated after-hours work. We present MDwAIstScheduler, a low- cost, belt-worn pipeline that lets a physician speak naturally dur- ing the encounter and have the resulting medications, allergies, labs/orders/referrals, follow-up scheduling, vitals, and problems land in the EHR as review-ready drafts. Building on our earlier prototype, which relied on cloud speech recognition and a cloud language model, the current pipeline runs both transcription and intent extraction entirely on-device. Using a medical-domain auto- matic speech recognition (ASR) model and a 1.7B-parameter lan- guage model we fine-tuned for clinical action extraction, no patient audio or text leaves the device, and the structured drafts are written directly into the Elation EHR for the physician to confirm. The result is a documentation tool that removes keyboard work from the visit without removing the clinician from the record, allowing them to focus on what matters most, patient care, while reducing burden at the same time.
