近三十年以来，恶性肿瘤在全球范围内已成为威胁人类健康最重要的疾病之一。目前，世界卫生组织公认的三大常规肿瘤治疗方法为手术治疗、放射治疗以及化疗，70％的恶性肿瘤患者适宜进行放射治疗，放射治疗的治愈率在肿瘤总体治愈率中占到18％。当前，高端医疗器械在临床诊疗活动中的不断应用，不可避免的会带来一些使用问题，而造成这些使用问题的主要诱因是由于医疗器械可用性设计不合格。作为治疗肿瘤的大型医疗设备，医用直线加速器在临床上的使用也不可避免的存在着较多的可用性问题。目前，国外针对放疗设备开展的可用性研究相对较少，国内关于放疗设备的可用性研究更是空白。本文从放射治疗流程中操作较为频繁的放疗控制软件着手，采用多模态测量手段，对放疗软件的可用性展开研究，为放疗软件优化改进提供理论支持和研究工具。本文的研究内容主要涵盖了以下几点： 第一，利用放疗设备可用性调研问卷，向43家放疗设备使用单位发放现状调研问卷，获得了5种品牌型号的临床在用放疗设备可用性现状调研数据。调研结果表明，我国临床在用放疗设备的可用性设计现状不容乐观，放疗设备可用性设计尚未达到用户预期水平；国产放疗设备与进口设备在可用性设计方面还存在差距，国产设备厂商需要加强可用性研究投入。此外，研究结果还表明放疗设备用户最为关注的可用性维度是设备的软件易用性。 第二，提出了一套基于多模态测量的放疗软件可用性研究测试方法，并分析了不同模态指标间的相关关系。通过被试在放疗软件上执行典型的日常操作任务，采集了被试在不同型号放疗软件完成任务时的各模态指标数据。测试结果表明，被试在不同放疗软件完成相同测试任务时，软件可用性设计差异主要体现在绩效(任务时间)、主观评价量表(任务难易程度、SUS量表、USE量表、NASA-TLX工作负荷量表)、眼动指标(瞳孔直径、平均注视时间、总注视时间、平均眨眼时间、眨眼频率)以及生理指标(皮肤导电性、皮肤温度)上。通过对四种不同模态的可用性评估指标采取相关性分析，发现它们之间存在着相关关系。因此，采用多模态测量方法更能全面评估被试对放疗软件可用性的直观感受。 第三，为放疗软件的可用性评估提供了预测模型。依据已有的放疗软件可用性研究数据，通过单因素方差分析确定了与放疗软件可用性评估相关的评价指标。采用支持向量回归机构建了放疗软件可用性评估预测模型。研究结果表明，综合使用绩效、主观评价量表、眼动、生理指标评价放疗软件可用性可获得最贴近实际的预测结果。 第四，探究了放疗软件主界面设计元素对可用性评估指标的影响程度，确定了影响放疗软件主界面可用性的关键设计元素。在已有的放疗软件可用性研究基础上，通过专家咨询法确定了放疗软件主界面5个重要设计元素以及每个设计元素两种不同的设计水平。采用正交试验研究方法，设计了八个不同类型的放疗软件主界面原型。基于多模态可用性研究测试，采集了被试完成测试任务时的相关实验数据。研究的结果表明，重要治疗参数显示对绩效指标任务完成时间具有显著性的影响；重要治疗参数的显示对NASA-TLX工作负荷量表得分有显著性的影响；联锁报警内容显示对眼动指标平均注视时间存在显著性影响。放疗软件主界面五个设计元素对被试的生理指标变化未产生显著性的影响。 通过本文的研究，为当前放疗设备的可用性现状提供了调研工具，为放疗软件可用性提供了测量方法，为放疗软件的可用性评估提供了预测模型，同时也为放疗软件的可用性优化改进提供理论支撑和研究工具。 关键词：放疗设备；放疗软件系统；多模态测量；可用性；支持向量回归机；预测评估

 In the past three decades, malignant tumor has become one of the most important diseases which threatening human health on a global scale. At present, the World Health Organization recognizes three major conventional cancer treatment methods: Surgical treatment, radiotherapy, and chemotherapy. Among them, the cure rate of the radiotherapy accounts for 18% of the total tumor cure rate, and tumor specialists have to apply the radiotherapy to at least the 70% of malignant tumor patients in the tumor treatment process. At present, the continuous application of high tech medical devices in clinical diagnosis and treatment activities will inevitably bring some usage problems, and the main cause of these usage problems is due to the unqualified design of the human-machine interface of medical devices. As the medical equipment for treating tumor, the clinical use of medical linear accelerator inevitably has many usability problems. At present, there are few relatively studies on the usability of radiotherapy equipment abroad, and the domestic research on the usability of radiotherapy equipment is nearly blank. This article starts with the most frequently used radiotherapy controlling software in the radiotherapy treatment process. This article uses multimodal measurement methods to study the usability of radiotherapy software, and provides with theoretical support and research tools for the optimization and improvement of radiotherapy software interactive interface. The research content of this paper mainly covers the following points: Firstly, through the survey questionnaire for the usability of radiotherapy equipment, we distributed it to 43 radiotherapy equipment application institutions. The survey data on the usability of clinical re-use radiotherapy equipment was obtained. The survey results show that the usability of clinical reusable radiotherapy equipment in China is not optimistic. The usability design of radiotherapy equipment has not yet reached the user's expectation; there is still a gap in the usability design between domestic radiotherapy equipment and imported equipment, and domestic equipment manufacturers need to strengthen the usability research investment. In addition, the results of the study also indicate that the usability dimension of the most popular users of radiotherapy equipment is the ease of use of the device. Secondly, a set of radiotherapy software usability research and test methods based on multimodal measurement was proposed, and the correlation between different modal indicators was analyzed. Through the typical daily operation tasks performed by the participants on the radiotherapy software, the modal data of the subjects in the different types of radiotherapy software were collected. The test results show that when participants completed the same test tasks in the different radiotherapy softwares, the software interaction interface usability design differences are mainly reflected in performance (task time), subjective evaluation scale (task difficulty level, SUS scale, USE scale, NASA-TLX workload scale), eye movement index (pupil diameter, average fixation time, total fixation time, average blink time, blink frequency) and physiological indicators (skin conductivity, skin temperature). Correlation analysis was carried out on the usability evaluation indicators of four different modes, and it was found that there was a correlation between them. Therefore, the multimodal measurement method can more fully assess the intuitive perception of the user's availability of radiotherapy software. Third, a predictive model is provided for the usability assessment of radiotherapy software. Based on the available radiotherapy software usability study data, the evaluation indicators related to the evaluation of the usability of radiotherapy software were determined by one-way analysis of variance. A support vector regression machine was used to construct a predictive model for the usability evaluation of radiotherapy software. The results of the study show that the comprehensive use performance, subjective evaluation scale, eye movement, physiological indicators to evaluate the interactive interface of radiotherapy software can obtain the most realistic prediction results. Fourth, the degree of influence of the main interface of radiotherapy software on the usability evaluation index was explored, and the key design factors affecting the usability of the main interface of the radiotherapy software were determined. Based on the existing research on the usability of radiotherapy software, five important design elements of the main interface of radiotherapy software and two different design levels of each design element were determined by expert consultation. Eight different types of radiotherapy software main interface prototypes were designed by orthogonal experimental research. Based on the multi-modal usability research test, the relevant experimental data when the participants completed the test task was collected. The results of the study showed that important treatment parameters showed a significant impact on the completion time of performance indicators; the display of important treatment parameters had a significant impact on the NASA-TLX workload scale score; the interlocking alarm content showed the eye movement index. There is a significant impact on the average gaze time. The five design elements of the main interface of the radiotherapy software did not have a significant impact on the changes in the physiological indicators of the subjects. Through the research in this paper, it provides with a research tool for the current usability of radiotherapy equipment, provides with a measurement method for the usability of the radiotherapy software interactive interface, provides with a predictive model for the usability evaluation of radiotherapy software, and also provides with optimization for the usability optimization of radiotherapy software. Therefore, the research results of this paper have strong guidance and application value for the institutions and equipment manufacturers of radiotherapy equipment. Key words: Radiotherapy equipment; radiotherapy software system; multimodal measurement; usability; support vector regression machine; predictive evaluation