Pituitary Microadenoma Diagnosis using Deep Learning

This repository contains the source code of PM-CAD system which is described in our paper "Automatic diagnosis of pituitary microadenoma from magnetic resonance imaging using deep learning algorithms". The PM-CAD system is a computer-aided diagnosis system aiming to diagnose pituitary microadenoma from MRI using deep learning techniques.

Dependencies

  • pytorch >= 1.6.0
  • SimpleITK >= 1.1.0
  • albumentations >= 0.4.3
  • tensorflow >= 2.0.0
  • numpy >=1.16.3
  • opencv-Python >= 4.1.2.30
  • matplotlib >= 2.2.2
  • tqdm >= 4.46.0

Preparation

Download the checkpoints from BaiduYunPan [Link] with code frnp.

Demo

We provide a snippet code to demonstrate how to diagnose pituitary microadenomas using our method, run by:

python demo.py --path=Dataset/Demo/pituitary_microadenoma_case

Train

To train the pituitary detection model:

cd DetectionModel
sh experiment/fasterrcnn/train_FasterRCNN.sh

To train the microadenomas diagnosis model (i.e., PM-Net):

cd DiagnosisModel
sh experiment/panet/run_PANet.py

To train other microadenomas diagnosis models:

cd DiagnosisModel
sh experiment/baseline/train_VGG16.sh
sh experiment/baseline/train_ResNeXt50.sh
sh experiment/baseline/train_ResNet50.sh
sh experiment/baseline/train_GoogLeNet.sh
sh experiment/baseline/train_DenseNet169.sh
sh experiment/cnn3d/train_CNN3D.sh