Early detection of cognitive impairment by natural language among outpatients and community dwellers
Early detection of cognitive impairment by natural language among outpatients and community dwellers
复制标题
在门诊患者和社区居民中，自然语言对认知障碍的早期发现
DOI：
10.1093/geroni/igaa057.3291
复制
发表时间：
2020-12-16
影响因子：
7
通讯作者：
Kamide K
中科院分区：
文献类型：
--
作者：
Kiyoshige E;Kabayama M;Takeya Y;Takami Y;Takeda S;Gondo Y;Rakugi H;Kamide K
背景：尽管早期发现认知能力下降与提高痴呆患者的生活质量有重要关系，但由于人们在发现认知障碍时通常需要进行神经心理测试，因此早期发现一直很困难。据报道，患者被告知的时间是认知能力下降更严重的时间，因此，我们旨在通过使用自然语言处理方法从简短的访谈中确定是否存在认知障碍。方法：本研究采用横断面分析方法，对老年门诊和社区居住老年人进行调查，调查对象包括七十岁、八十岁、九十岁和百岁老人。认知衰退通过日本人认知状态电话访谈（tic - j）进行评估，并以二元结果建模（截止值<33点）。通过半结构化访谈收集自然语言数据，了解健康状况和空间、时间和地点的认知取向。我们使用开源文本分割库将自然语言文本解析为词袋和词频率逆文档频率（TF-IDF）表示。结果：门诊患者38例（19.9%），社区居民153例（80.1%），认知功能障碍60例（31.4%）。时间认知取向的TF-IDF得分最高为0.49。在这个问题中，没有认知障碍的参与者无法计算得分。在有和没有认知障碍的参与者之间，TF-IDF得分没有显著差异。结论：无认知障碍的老年人可能没有及时发生认知取向发作，这可能有助于早期发现认知障碍
Background: Although early detection of cognitive decline has a significant relation to improving the quality of life of dementia patients, this early detection has been difficult due to requires of neuropsychological tests which people generally take when they notice their cognitive impairment. The timing of patients’ notice was reported to be worse cognitive decline already, thus, we aimed to determine if cognitive impairment from a short interview by using Natural Language Processing approach. Methods: The present study used cross-sectional analysis among elderly outpatients and community-dwelling elderly from Septuagenarians, Octogenarians, Nonagenarians Investigation with Centenarians (SONIC) study. Cognitive decline was assessed by Telephone Interview of Cognitive Status for Japanese (TICS-J) and modeled as a binary outcome (cut-off <33 points). Natural language data was collected by semistructured interviews about health conditions and cognitive orientation in space, time, and place. We used an open-source text segmentation library to parse natural language text into bag-of-words and term frequency-inverse document frequency (TF-IDF) representations. Results: There were 38 (19.9%) outpatients and 153 (80.1%) community dwellers, and 60 (31.4%) participants were defined as cognitive impairment. The maximized TF-IDF score was 0.49 in cognitive orientation in time questions. In this question, participants without cognitive impairment could not calculate the score. There were no significant differences in TF-IDF scores between participants with and without cognitive impairment. Conclusions: Elderly without cognitive impairment might not have an episode about cognitive orientation in time, and this may help for early detection of cognitive impairment