sponsor & contact us

A. 高雄醫學大學附設醫院麻醉部
T. 07-3121101 #7033、7035
F. 07-3217874
E. 2020annualmeetingtsa@gmail.com
陳冠甫 Kuan-Fu Chen
陳冠甫Kuan-Fu Chen
個人簡介

現職
長庚大學臨床資訊醫學統計中心副教授  
長庚紀念醫院醫師研究員
基隆長庚紀念醫院急診醫學部
副教授級主治醫師
台灣急診醫學會編輯委員會副主任委員


個人經歷
美國約翰霍普金斯大學急診醫學部研修醫師
林口長庚紀念醫院急診醫學部總住院醫師

人工智慧在急重症醫療的應用:發展與陷阱

在急重症醫療之中,臨床預測模型廣泛使用來為病人的疾病診斷、預後、以及治療效果做應用。預測模型的建立遠在機器學習及深度學習盛行之前便是一門學問。隨著近年來各式各樣的工具問市,機器學習及深度學習成為不再令人望而生畏的建模方法。然而,許多藏在細節裡的魔鬼,並不會因為這些工具問市而現出原形。第一線臨床醫療人員面對數以千計新問市的預測模型,往往沒有足夠的知識評估運用的可行性。

在這短短半小時的分享裡,主講人將從資料前處理、遺失值處理、研究設計、特徵挑選、各式機器模型挑選、過度配適、時間序列設置、到深度學習層次設定等等的細節,配合身為急重症醫師的臨床資料,給聽眾做一次性的醍醐灌頂式的分享,希望不論是有意在未來從事人工智慧相關研究,或是希冀了解相關預測模型的評讀,都有基本的收獲。



 

Application of Artificial Intelligence in Emergent and Critical Care Medicine: Developing Trend and Caveat

Clinical prediction models have been widely utilized for disease diagnoses, prognoses, and effectiveness of treatment in emergent and critical care medicine. The development of clinical prediction models has been specialized before the availability of machine learning and deep learning. With lots of friendly tools developed for these purposes, machine learning and deep learning have become more possible for researchers. However, the devils in the details would not disappear because of these tools. Front-line health care providers often do not have sufficient knowledge to appraise these models.

In this half an hour talk, the speakers will describe the detail of data preprocessing, missing data management, study design, feature selection, model selections, overfitting, time-series setting, to setting of the deep learning model. We hope the audience could obtain enough knowledge no matter they would like to participate in artificial intelligence research or to deploy the models to their clinical work.