人工智能臨床病史診斷失智程度與類型問卷的初步成效

Preliminary Study of the History-based Artificial Intelligent Clinical Dementia Diagnostic System

目的:建立一個臨床病史診斷失智症的問卷,以運用在判斷疾病類別與嚴重度的診斷系統上面,並且可以藉由人工智能與機器學習以增強診斷的準確度。 方法:在台灣兩個中心同時蒐集失智症病患,每位受試者接受一份標準化的結構式臨床病史問卷。我們比對了受試者的單純的問卷診斷與最後診斷的異同。 結果:全部共 104 位受試者,在與確定診斷比較之下,單純問卷診斷準確率分別為:阿茲海默症 94.6%、血管性失智症 73.3%、路易氏體失智症 92.6% 與巴 金森病失智症 89.5%。而失智程度與臨床失智量表比較下,也是有相當高的吻合度。結論:我們的研究顯示,此問卷對於常見的失智症的疾病診斷與疾病嚴重度都有相 當好的輔助判斷能力,因此可運用在往後的流行病學研究或是前瞻性研究。

 

Aim and purpose: The aim of this study is to establish a history-based clinical diagnostic system that can be applied to the clinical diagnosis of dementia severity and subtypes. Furthermore, the diagnostic accuracy can be increased via artificial intelligence and machine learning process. Methods: We consecutively enrolled individuals with normal cognition, mild cognitive impairment, or dementia from 3 centers in Taiwan. All of them received a structured history-based questionnaire and a complete dementia survey. We further compared the consistency of diagnosis according to the questionnaire and the final diagnosis made by complete dementia survey.Results: The diagnoses of a total of 104 individuals were compared. The consistency of the diagnoses of subtypes of dementia were 94.6% for Alzheimer’s disease,73.3% for vascular dementia, 92.6% for dementia with Lewy bodies, and 89.5% for Parkinson’s disease dementia. The diagnosis of severity of dementia was also highly compatible with the diagnosis according to the Clinical Dementia Rating Scale.Conclusion: This result demonstrated the satisfied diagnostic accuracy and consistency of the history-based dementia questionnaire. The screening tool is suitable to be applied to further prospective studies as well as populational studies.