3DConvNet Solutions for Medical Image Challenges

This repository contains 3d ConvNet Solutions for Medical Image Challenges.
The project is based on Tencent MedicalNet and MONAI framework
which provides a series of 3D-ResNet pre-trained models and domain-optimized foundational capabilities for developing healthcare imaging training workflows.

Update(2020/05/01)

I provide a baseline 3DConvNet code for TReNDS Neuroimaging challenge host on Kaggle.

Requirements
Code
Experiments
TODO

Requirements

Python 3.7.0
PyTorch-1.5
monai-0.1.0

Code

Structure of data directories base on MedicalNet

/
    |--datasets/：Data preprocessing module
    |   |--brains18.py：MRBrainS18 data preprocessing script
    |   |--RSNA19.py：RSNA19 data preprocessing script
    |   |--TReNDs.py：TReNDs data preprocessing script
    |--models/：Model construction module
    |   |--resnet.py：3D-ResNet network build script
    |--utils/：tools
    |   |--logger.py：Logging script
    |--toy_data/：For CI test
    |--data/：Data storage module
    |   |--MRBrainS18/：MRBrainS18 dataset
    |   |   |--images/：source image named with patient ID
    |   |   |--labels/：mask named with patient ID
    |   |--train.txt: training data lists
    |   |--val.txt: validation data lists
    |--pretrain/：Pre-trained models storage module
    |--model.py: Network processing script
    |--setting.py: Parameter setting script
    |--train_MRBrainS18.py: MRBrainS18 training demo script
    |--train_TReNDs.py: TReNDs training script
    |--train_RSNA19.py
    |--README.md

Download data & pre-trained models from Tencent MedicalNet official repo (Google Drive or Tencent Weiyun)

Network structure parameter settings

Model name   : parameters settings
resnet_10.pth: --model resnet --model_depth 10 --resnet_shortcut B
resnet_18.pth: --model resnet --model_depth 18 --resnet_shortcut A
resnet_34.pth: --model resnet --model_depth 34 --resnet_shortcut A
resnet_50.pth: --model resnet --model_depth 50 --resnet_shortcut B
resnet_101.pth: --model resnet --model_depth 101 --resnet_shortcut B
resnet_152.pth: --model resnet --model_depth 152 --resnet_shortcut B
resnet_200.pth: --model resnet --model_depth 200 --resnet_shortcut B

Baseline for TReNDS Neuroimaging challenge

3D-Resnet10 trained from scratch Pretrained models

Network	fold 0	fold 1	fold 2	fold 3	fold 4
3D-Resnet10 Train from scratch	0.1700	0.1685	0.1729	0.1734	0.1734
3D-Resnet10 MedicalNet pretrained	0.1694	0.1691	0.1726	0.1746	0.1734

Computational Cost

GPU：NVIDIA Tesla P40

Network	Paramerers (M)	Running time (s)
3D-ResNet10	14.36	0.18
3D-ResNet18	32.99	0.19
3D-ResNet34	63.31	0.22
3D-ResNet50	46.21	0.21
3D-ResNet101	85.31	0.29
3D-ResNet152	117.51	0.34
3D-ResNet200	126.74	0.45

Please refer to Med3D: Transfer Learning for 3D Medical Image Analysis for more details：

TODO

Baseline (pure 3D ConvNet) code for TReNDS Neuroimaging challenge
Code and pretrained models for Intracranial-Hemorrhage-Detection (RSNA2019 challenge dataset)
More baseline code and pretrained models for recent Medical Image Challenges

125 lines (111 with data), 4.0 kB