dwilib

NEW: Switched to Python version 3, it is now required and assumed everywhere!

Tools in development to support analysis of Diffusion-Weighted Imaging (DWI) data, with focus on prostate cancer.

This software is being developed as part of a research project at the Magnetic Resonance Imaging Research Center, Turku, Finland.

Note: This is very much under construction. The program code is being written on demand basis, for custom needs at in-house projects. It has grown organically at the same time as the programmer has been studying the subject and learning the tools. So the code is quite messy in many places.

Features

• Read input data as DICOM, or in custom ASCII or MATLAB formats
• Perform model fitting (Monoexponential ADC, Kurtosis, Stretched exponential, Biexponential)
• Calculate correlation with Gleason score
• Calculate and compare diagnostic ROC AUCs
• Calculate reproducibility measures
• Plotting schemes
• Viewer for multi-slice, multi-b-value DWI DICOM files (uses the Matplotlib GUI widget)
• Build tool for automated pipelining of data processing tasks

Todo or In Progress

• Improved documentation
• Improved build tool
• Autonomous tumor delineation/ROI placement
• Regression classification

Requirements

• Python 3.4
• pathlib2 (or newer Python)
• NumPy
• SciPy
• Scikit-Learn

Optional requirements

• leastsqbound-scipy (fitting)
• Matplotlib (plotting)
• Pydicom (reading DICOM files)
• h5py (handling HDF5 files)
• scikit-image (texture analysis)
• Mahotas (texture analysis)
• PyDoIt (task management)