Automated Grading of Peripheral Facial Palsy

This repository contains the underlying data of the thesis "Automated Grading of Facial Palsy" by T.C. ten Harkel. These data are separated in the following 4 folders: Chapter 2, Chapter 3 (part 1 & 2), Chapter 4, and Chapter 5. Each chapter folder contains a separate Readme.txt file giving a detailed description of the data contained in that folder. A brief overview can be found below:

Chapter 2 :
This folder contains the RealSense F200 depth accuracy (mm) for each individual patient for each pose of the Sunnybrook Facial Grading System (SFGS).

Chapter 3 part 1:
This folder contains the following data: (1) the landmarks as placed by the observers for each individual patient during the neutral pose of the SFGS, (2) the depth accuracy of the RealSense D415 for each individual patient during the neutral pose of the SFGS, (3) the landmark placement accuracy between the observers for each individual patient during the neutral pose of the SFGS, (4) the difference between the anthropometric measurements between observers for each individual patient during the neutral pose of the SFGS.

Chapter 3 part 2:
This folder contains the following data: (1) the landmarks as placed by the observers for each individual patient during the voluntary movements of the SFGS, (2) the depth accuracy of the RealSense D415 for each individual patient during the voluntary movements of the SFGS, (3) the landmark placement accuracy between the observers for each individual patient during the voluntary movements of the SFGS, (4) the difference between the anthropometric measurements between observers for each individual patient during the voluntary movements of the SFGS.

Chapter 4:
This folder contains the actual SFGS score and predicted SFGS score of the CNN model for all individual patients, divided into the 13 separate SFGS elements and five stratified k-folds.

Chapter 5:
This folder contains the actual SFGS score and predicted SFGS score of the CNN model for all individual patients, divided into the 13 separate SFGS elements and five stratified k-folds.

本仓库收录了T.C. ten Harkel所著学位论文《面神经麻痹自动化分级》（Automated Grading of Facial Palsy）的配套基础数据集。本数据集划分为以下4个文件夹：第2章、第3章（第1、2部分）、第4章及第5章。每个章节文件夹内均附带独立的Readme.txt文件，对该文件夹内包含的数据进行详细说明，下文将提供简要概览：

### 第2章
本文件夹包含每位患者在桑德布鲁克面部分级系统（Sunnybrook Facial Grading System, SFGS）各姿态下的RealSense F200深度精度（单位：毫米）。

### 第3章第1部分
本文件夹包含以下四类数据：(1) 观察者在SFGS静态中立姿态下为每位患者标注的面部地标点；(2) 每位患者在SFGS静态中立姿态下的RealSense D415深度精度；(3) 观察者间在SFGS静态中立姿态下的面部地标点定位精度；(4) 观察者间在SFGS静态中立姿态下的人体测量学测量值差值。

### 第3章第2部分
本文件夹包含以下四类数据：(1) 观察者在SFGS自主运动姿态下为每位患者标注的面部地标点；(2) 每位患者在SFGS自主运动姿态下的RealSense D415深度精度；(3) 观察者间在SFGS自主运动姿态下的面部地标点定位精度；(4) 观察者间在SFGS自主运动姿态下的人体测量学测量值差值。

### 第4章
本文件夹收录了所有患者的SFGS真实评分与卷积神经网络（Convolutional Neural Network, CNN）模型预测的SFGS评分，数据按SFGS的13个独立分级要素以及5层分层k折交叉验证进行划分。

### 第5章
本文件夹收录了所有患者的SFGS真实评分与卷积神经网络（Convolutional Neural Network, CNN）模型预测的SFGS评分，数据按SFGS的13个独立分级要素以及5层分层k折交叉验证进行划分。