junqiangchen / pytorchdeeplearing Goto Github PK

View Code? Open in Web Editor NEW

167.0 5.0 32.0 8.08 MB

Meidcal Image Segmentation Pytorch Version

Python 100.00%

classification-model pytorch-implementation segmentation-models unet-image-segmentation vnet3d flask flask-application

pytorchdeeplearing's Introduction

PytorchDeepLearing

ImageSegment Model

There are some Model NetWorks of 3D ImageSegment and 2D ImageSegment

ImageSegment Nets

There are Unet and Vnet Family all has 2d and 3d version.

ImageSegment Loss Function

There are some loss functions of 3D ImageSegment and 2D ImageSegment

How to Use

i have reimplemented the image segmentation loss functions with pytorch1.10.0

there are binary_crossentropy,dice_loss,focal_loss_sigmod etc all has 2d and 3d version.

there are categorical loss functions of crossentropy,dice_loss,focal_loss etc all has 2d and 3d version.

MS-SSIM loss and SSIM loss for calculating image similarity.

centerline dice loss for vessel segmentation

there are 9 type of segment metric,including dice,surface disatance,jaccard,VOE,RVD,FNR,FPR,ASSD,RMSD,MSD,etc.

flask_app.py is the demo example of the Flask Deep Learning Segmentation Model Service Deployment.

Contact

https://github.com/junqiangchen
email: [email protected]
WeChat Public number: 最新医学影像技术

pytorchdeeplearing's People

Contributors

Stargazers

Watchers

pytorchdeeplearing's Issues

LUNA22-肺结节分类挑战

看到您LUNA22-肺结节分类挑战的文章，觉得很有意思，文章里说您将代码放在了这里，可我没找到，请问您LUNA22-肺结节分类挑战具体的代码在哪里呢

ROI区域代码请教

由于结节在图像中心位置，且目标比较小，所以根据LIDC-IDRI_1176.npy文件中结节的直径大小，在128x128x64大小上获取肺结节ROI区域图像，这样就可以进一步减少背景干扰，因此得到肺结节的ROI区域。然后统计全部肺结节ROI区域的大小，平均大小是33x33x18。
请问这里获取ROI区域的代码能分享一下吗

你好，flare2022训练集中2000例未标注的数据是怎么处理的

您好，请问《ASOCA2020——冠状动脉分割》这个的数据集或者您的训练模型能分享一下吗

您好，请问《ASOCA2020——冠状动脉分割》这个的数据集或者您的训练模型能分享一下吗？谢谢！

青光眼那个用的是BinaryResNet2dModel？

这个配置么？TASK1

resnet2d = BinaryResNet2dModel(image_height=1024, image_width=1024, image_channel=1, numclass=1,
                              batch_size=32, loss_name='BinaryCrossEntropyLoss',use_cuda=use_cuda)
resnet2d.trainprocess(trainimages, trainlabels, valimages, vallabels, model_dir='log/BinaryResNet2d/CE', epochs=300,
                      lr=0.005)

有关luna2022的代码请教

大神您好，我想请教一下关于从zenodo下载的数据集怎么处理啊？里面是nii型文件，而您的instruction相关的是npy文件，不是太懂这里面的关系。

我看到您的github的 dataprocess 的文件下有个data里面是储存的csv，我看了一些数据处理的代码不太能理解，关于如何从数据集nii到npy到train，validation分类这个步骤不是很懂。能否麻烦大神您给个详细点的步骤。

如果方便的话能麻烦您写一个luna2022数据集分类的代码执行步骤吗，我是初学者，根据您在GitHub上的代码感觉到很难理解