Please click HERE to DOWNLOAD the model.
This code is packaged and released. Users can unzip release_model.zip file, and double click ‘HCC_moedel.exe’ to run the program.
Instructions：
1. Double click ‘HCC_model.exe’.
2. Click ‘browse’ button, and input the original image folder path “Dicom file path”, segmentation mask folder path “Segment file path”, and clinical table file path “Clinic info file path”. An example of clinical table, see download file “Clinical_info.xlsx”.
3. After inputting the file paths, click ‘start’ button. Wait for less than 30-40 seconds, the predicted status of MaVI of the patient can be acquired, including probability of MaVI, future MaVI occurrence status, survival risks, etc.
Note:
1) The original image format should be medical imaging format ‘.dicom’. Original images modality includes CT and MR.
2) The segmentation mask format should be medical imaging format ‘.nrrd’. Users could implement segmentation by ITK-SNAP, 3D slicer, etc.
3) Clinical variable only conclude node-in-node sign. 0 represents for no node-in-node sign; 1 represents for node-in-node sign existing. Clinical table example can be acquired in the download file “Clinical_info.xlsx”.
4) If it reports error when opening the program, please close antivirus software first.
5) It may cost 1-2 minutes to open the program for the first time use.
This software is for ACADEMIC USE only.
Please cite article “Noninvasive prediction of subsequent macrovascular invasion in hepatocellular carcinoma by machine learning-based imaging analysis”, if using the model and program.
Implemented by Jingwei Wei, Dongsheng Gu, and Xiaohan Hao. Please contact weijingwei2014@ia.ac.cn or dongshegngu2016@ia.ac.cn, if you have any question.
本代码打包发布，用户可通过解压release_model.zip 文件，双击release_model文件夹下的‘HCC_model.exe’文件运行程序。
操作说明：
1、双击HCC_model.exe文件,出现窗口界面。
2、按图中提示，点击‘browse’按钮，输入: 原始文件路径“Dicom file path”, 分割文件路径“Segment file path”, 临床信息量表路径“Clinic info file path”。临床信息量表样例请见下载文件“Clinical_info.xlsx”。
3、路径选择完毕后，点击start按钮，等待一段时间（一般不超过一分钟），即可得到该病人的预测信息，包括：大学管侵犯预测概率、预测状态，血管侵犯时间风险、无进展生存风险、总生存风险。
注意：
1） 病人原始文件格式须为dicom医学图像格式，支持CT，MR等多种原始文件序列；
2） 临床信息表格只需要结节信息，0代表无结节征，1代表有结节征，表格例程在文件夹下xlsx文件；
3） 临床信息表格只需要结节信息，0代表无结节征，1代表有结节征，表格例程在文件夹下clinical_Info.xlsx文件；
4） 运行时如有杀毒软件报错，请关闭杀毒软件；
5） 初次使用，打开软件过程或耗费1-2分钟。
此软件仅作为学术性软件使用，非商用软件！中国科学院自动化研究所分子影像重点实验室保留一切相关法律权利。
如使用模型或程序，请引用文章： “Noninvasive prediction of subsequent macrovascular invasion in hepatocellular carcinoma by machine learning-based imaging analysis”。
本程序由魏靖伟、顾东升及郝小涵开发实现。如有问题，请联系weijignwei2014@ia.ac.cn或gudongsheng2016@ia.ac.cn。