Taveekitworachai Pittawat, Chanmas Gunt, Paliyawan Pujana, Thawonmas Ramita, Nukoolkit Chakarida, Dajpratham Piyapat, Thawonmas Ruck
Graduate School of Information Science and Engineering, Ritsumeikan University, 2-150 Iwakura-cho, Ibaraki, 567-8570, Osaka, Japan.
Ritsumeikan Center for Game Studies, Ritsumeikan University, 56-1 Toji-in Kitamachi, Kita, 603-8577, Kyoto, Japan.
Heliyon. 2024 Jun 17;10(12):e32930. doi: 10.1016/j.heliyon.2024.e32930. eCollection 2024 Jun 30.
Simulator-based driving assessments (SA) have recently been used and studied for various purposes, particularly for post-stroke patients. Automating such assessment has potential benefits especially on reducing financial cost and time. Nevertheless, there currently exists no clear guideline on assessment techniques and metrics available for SA for post-stroke patients. Therefore, this systematic review is conducted to explore such techniques and establish guidelines for evaluation metrics. This review aims to find: (a) major evaluation metrics for automatic SA in post-stroke patients and (b) assessment inputs and techniques for such evaluation metrics. The study follows the PRISMA guideline. Systematic searches were performed on PubMed, Web of Science, ScienceDirect, ACM Digital Library, and IEEE Xplore Digital Library for articles published from January 1, 2010, to December 31, 2023. This review targeted journal articles written in English about automatic performance assessment of simulator-based driving by post-stroke patients. A narrative synthesis was provided for the included studies. The review included six articles with a total of 239 participants. Across all of the included studies, we discovered 49 distinct assessment inputs. Threshold-based, machine-learning-based, and driving simulator calculation approaches are three primary types of assessment techniques and evaluation metrics identified in the review. Most studies incorporated more than one type of input, indicating the importance of a comprehensive evaluation of driving abilities. Threshold-based techniques and metrics were the most commonly used in all studies, likely due to their simplicity. An existing relevant review also highlighted the limited number of studies in this area, underscoring the need for further research to establish the validity and effectiveness of simulator-based automatic assessment of driving (SAAD). More studies should be conducted on various aspects of SAAD to explore and validate this type of assessment.
基于模拟器的驾驶评估（SA）最近已被用于各种目的并得到研究，特别是针对中风后患者。使这种评估自动化具有潜在益处，尤其是在降低财务成本和时间方面。然而，目前尚无针对中风后患者SA的评估技术和指标的明确指南。因此，进行了这项系统综述，以探索此类技术并建立评估指标指南。本综述旨在找到：（a）中风后患者自动SA的主要评估指标，以及（b）此类评估指标的评估输入和技术。该研究遵循PRISMA指南。在PubMed、科学网、ScienceDirect、ACM数字图书馆和IEEE Xplore数字图书馆上进行了系统检索，以查找2010年1月1日至2023年12月31日发表的文章。本综述针对用英文撰写的关于中风后患者基于模拟器驾驶的自动性能评估的期刊文章。对纳入的研究进行了叙述性综合。该综述纳入了6篇文章，共有239名参与者。在所有纳入的研究中，我们发现了49种不同的评估输入。基于阈值、基于机器学习和驾驶模拟器计算方法是综述中确定的三种主要评估技术和评估指标类型。大多数研究纳入了不止一种类型的输入，这表明全面评估驾驶能力的重要性。基于阈值的技术和指标在所有研究中使用最为普遍，可能是因为它们简单易行。一项现有的相关综述也强调了该领域研究数量有限，强调需要进一步研究以确立基于模拟器的驾驶自动评估（SAAD）的有效性和效能。应该对SAAD的各个方面进行更多研究，以探索和验证这种类型的评估。
