dvha-stats 0.1.7

Simple DICOM tag editor built with wxPython and pydicom

A library of prediction and statistical process control tools. Although based on work in DVH Analytics, all statistical tools in this library are generic and not specific to radiation oncology. See our documentation for advanced uses.

What does it do?

• Read data from CSV or supply as numpy array

• Basic plotting

  • Simple one-variable plots from data

  • Control Charts (Univariate and Multivariate)

  • Heat Maps (correlations, PCA, etc.)

• Perform Box-Cox transformations

• Calculate Correlation matrices

• Perform Multi-Variable Linear Regressions

• Perform Principal Component Analysis (PCA)

Other information

• Free software: MIT license

• Documentation: Read the docs

• Tested on Python 3.6, 3.7, 3.8

Dependencies

• scipy

• numpy

• scikit-learn

• regressors

• matplotlib

Basic Usage

from dvhastats.ui import DVHAStats
s = DVHAStats("tests/testdata/multivariate_data.csv")

>>> s.var_names
['V1', 'V2', 'V3', 'V4', 'V5', 'V6']

>>> s.show('V1')  # or s.show(0), can provide index or var_name

>>> s.show(0, plot_type="hist")

Pearson-R Correlation Matrix

pearson_mat = s.correlation_matrix()
>>> pearson_mat.show()

Spearman Correlation Matrix

spearman_mat = s.correlation_matrix("Spearman")
>>> spearman_mat.show()

Univariate Control Chart

ucc = s.univariate_control_charts()
>>> ucc["V1"].show()  # or ucc[0].show(), can provide index or var_name

Multivariate Control Chart

ht2 = s.hotelling_t2()
>>> ht2.show()

Multivariate Control Chart (w/ non-normal data)

ht2_bc = s.hotelling_t2(box_cox=True)
>>> ht2_bc.show()

Multi-Variable Linear Regression

import numpy as np
y = np.random.rand(s.observations)
mvr = s.linear_reg(y)
>>> mvr.show()

>>> mvr.show("prob")

Principal Component Analysis (PCA)

pca = s.pca()
>>> pca.show()