GBM_PCNSL_CNN_modeling

Deep Learning for Automatic Differential Diagnosis of Primary Central Nervous System Lymphoma and Glioblastoma: Multi-parametric Magnetic Resonance Imaging based Convolutional Neural Network Model

The schematic workflow of CNN models development. The first step consisted of a series of automatic image preprocessing procedures, including image registration, brain extraction, and standardization. After image preprocessing, DenseNet based CNN models were developed by transfer learning, and a single-parametric CNN model was built using CE-T1WI, FLAIR, and ADC, respectively. A multi-parametric CNN model was developed using image-level fusion and decision-level fusion, respectively.

More details, please refer to our paper:
Wei Xia, Bin Hu, Haiqing Li, Wei Shi, Ying Tang, Yang Yu, Chen Geng, Qiuwen Wu, Liqin Yang, Zekuan Yu, Daoying Geng, Yuxin Li.
Deep Learning for Automatic Differential Diagnosis of Primary Central Nervous System Lymphoma and Glioblastoma: Multi-parametric Magnetic Resonance Imaging based Convolutional Neural Network Model, Journal of Magnetic Resonance Imaging, 2021. (https://doi.org/10.1002/jmri.27592)

code structure

CNN_modeling/: CNN model training, test and a few sample images.

CNN_training_program/: CNN model training, heat map generation, and a link for sample images. CNN_model_test_and_statistical_analysis/: CNN model test and performance metrics calculation.

radiomics_modeling/: radiomics features extraction, feature selection, and model building.

preprocessing/: Executable programs and scripts for image preprocessing (DicomSeriesReadImageWrite: dicom series read and volume write, bet2: brain mask extraction, Resampler: resampling to the reference volume).

statistical_analysis/: Radiologists' performance metrics calculation and statistical analysis.