AloNet Supplementary Repository

AloNet Author: Solam Lee ([e-mail]) AloNet is a convolutional neural network based on U-Net that can identify the hair loss and the scalp area by analying clinical photograph. This model was developed for the automated calculation of the Severity of Alopecia Tools (SALT) score in assessment of patients with alopecia areata. This repository posts the Mendeley Supplementary Materials, the program code, and the relevant data used in the paper titled "Clinically Applicable Deep Learning Framework for Measurement of the Severity of Alopecia Tool Score in Patients with Alopecia Areata". Along with the programs in the "/Program/" directory, a total of 2716 pixelwise annotations used for train the hair loss identifier (mask) and the hair loss identifier (target) could be find in the "/Data/" directory. However, please note that the clinical photograph of the patients could not be made publicly available because of strict privacy regulation. Before using AloNet program with your dataset, you should convert your dataset into numpy files. One clinical photograph (saved in .jpg with RGB format) need each annotation for the scalp area (saved in .gif with black&white color) and the hair loss (saved in .gif with black&white color), respectively. Please make sure that they have same image size each other, or the conversion will fail. We are now currently working on several postprocessing algorithms for AloNet to be available for general use. The Flask web application and its code will be made available publicly when the program is ready to use.AloNet 作者：Solam Lee（[e-mail]） AloNet是一款基于U-Net的卷积神经网络，可通过分析临床照片识别脱发区域与头皮区域。本模型专为斑秃患者病情评估中的脱发严重程度工具（Severity of Alopecia Tools，SALT）评分自动化计算而开发。 本仓库发布了题为"Clinically Applicable Deep Learning Framework for Measurement of the Severity of Alopecia Tool Score in Patients with Alopecia Areata"的论文所附带的Mendeley补充材料、程序代码及相关数据集。 除`/Program/`目录下的程序外，`/Data/`目录中还包含总计2716组用于训练脱发识别器（掩码）与脱发识别器（目标标签）的逐像素标注数据。但需注意，由于严格的隐私法规限制，患者的临床照片无法公开获取。 在使用AloNet程序处理您的数据集前，请先将数据集转换为numpy文件格式。每一张以RGB格式存储为JPG（.jpg）的临床照片，需分别对应一组头皮区域标注（以黑白GIF（.gif）格式存储）与一组脱发区域标注（以黑白GIF（.gif）格式存储）。请确保所有文件的图像尺寸一致，否则将导致转换失败。 目前团队正在针对AloNet开发多款后处理算法，以使其可通用化。待程序就绪后，基于Flask的Web应用及其源代码将对外公开。

搜集汇总

背景与挑战

背景概述

该数据集为'AloNet Supplementary Repository'，是一个医学图像分析数据集，基于U-Net的卷积神经网络AloNet，用于自动计算斑秃患者的SALT评分。它包含2716个像素级注释用于训练脱发识别模型，但原始临床照片因隐私问题未公开；同时提供程序代码和补充材料，并计划未来发布后处理算法和Web应用。