唐氏症候群在於各國的患者面容有相似的特徵，例如：面部起伏較小，鼻梁塌陷，
張口吐舌和耳格骨發育異常等等；但是沒有特定的表徵一定會出現在每一位唐氏症患者
身上。本文探討如何利用臉部影像資料，以及常用的機器學習方法來辨識唐氏症患者。
我們收集唐氏症患者與非唐氏症人員之照片，將臉部影像資料經過資料處理後，找出五
官之比例，再使用相關的資料探勘方法，如邏輯斯迴歸(Logistic regression)、交叉驗證
(Cross-validation) 及ROC 曲線(receiver operating characteristic curve) 等，並參考醫生
的經驗法則，期望能找出最有效之影像辨識分類方法來建立影像中唐氏症病徵的主要規
則。

There are several similar facial features in Down’s syndrome patients, such as a slightly
flattened facial profile, at nasal bridge, protruding tongue, abnormal outer ears and so
on; but not all the features would appear on every Down’s syndrome. The purpose of
this study was to discuss how to identify Down’s syndrome by using the facial image data
and statistical learning methodology. Based on photos of Down’s syndrome patients and
non-Down’s syndrome , we transform the facial image data into form such as proportion of
feature . Later, we used related statistical classification methods such as logistic regression,
cross- validation, receiver operating characteristic curve and so on, and referred to the rule
of thumb for doctors to clarify the status of as image. We tried to find the most effective
logistic model for classification and connect to the rules of Down’s syndrome symptoms.

論文審定書+i
誌謝+ii
摘要+iii
Abstract+ iv
1 前言+1
2 研究簡介+1
2.1 說明與基本統計量+1
2.2 資料處理+2
2.3 解釋變數+6
3 研究方法+9
3.1 逐步迴歸+9
3.2 邏輯斯迴歸+9
3.3 交叉驗證+10
3.4 ROC 曲線+10
4 研究結果+13
4.1 利用眼睛、鼻子與臉型長度黃金比例影像辨識結果+13
4.2 眼睛與眼睛間距及臉型長度黃金比例影像辨識結果+17
5 結論與建議+20
參考文獻+22
附錄+23

[1] Agresti A., (2002). Categorical Data Analysis,2nd, John Wiley & Sons, Inc., Hoboken,
New Jersey.
[2] Kohavi, R., (1995). A study of cross-validation and bootstrap for accuracy estimation
and model selection. Proceedings of the Fourteenth International Joint Conference on
Artificial Intelligence , 2, 1137-1143.
[3] Kutner M. H., Nachtsheim C. J. and Neter J.,(2004). Applied Linear Regression Models,
4th, McGraw-Hill, New York.
[4] Pentland E. A., (1994). View-based and modular eigenspaces for face recognition.
M.I.T Media Laboratory Perceptual Computing Section Technical Report, 245.
[5] Sung K. K. and Poggio T., (1994). Example-based learning for view-based human face
detection. C.B.C.L. Paper , 112.
[6] Turk M. and Pentland A., (1991). Pentland Eigenfaces for recognition.
J. Cognitive Neuroscience, 1, 71-86.
[7] 吳明衛(2003)，自動化臉部表情分析系統，國立成功大學資訊工程學系碩士論文。
[8] Down syndrome from Wikipedia, http://en.wikipedia.org/wiki/Down syndrome
[9] GAP, http://www.hmwu.idv.tw/gapsoftware/
23