pySuStaIn

SuStaIn algorithm in Python, with a combination of mixture style (i.e. EBM style) and z-score style SuStaIn implementations.

SuStaIn papers

• Young et al. 2018

Dependencies

• Python >= 3.5
• NumPy >= 1.18
• SciPy
• Matplotlib
• Scikit-learn for cross-validation
• kde_ebm for mixture modelling (KDE and GMM included)
• pathos for parallelization

Parallelisation

• Added parallelized startpoints

Running different SuStaIn implementations

In simrun.py sustainType can be set to:

• "mixture_GMM" : mixture model style SuStaIn with Gaussian mixture modelling of normal/abnormal.
• "mixture_KDE": mixture model style SuStaIn with Kernel Density Estimation (KDE) mixture modelling of normal/abnormal.
• "zscore": z-score style SuStaIn with three events for each biomarker (1,2,3 std. devs. from normality)