ReadMe.txt 2.82 KB
Newer Older
Jens Behrmann's avatar
Jens Behrmann committed
1 2 3 4 5 6 7 8 9 10 11 12
##########################################################
ReadMe for the code of the article "Deep Learning for 
Tumor Classification in Imaging Mass Spectrometry" by
Jens Behrmann, Christian Etmann, Tobias Boskamp,
Rita Casadonte, Jörg Kriegsmann, Peter Maass.

Should any questions arise, we would be happy to help you:

jensb@math.uni-bremen.de
cetmann@math.uni-bremen.de

The data is available at:
13
 https://seafile.zfn.uni-bremen.de/d/334b30f1a2894e0c8634/
Jens Behrmann's avatar
Jens Behrmann committed
14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

##########################################################

The provided files are structured in the following way:

1) CNNs_Bioinformatics

	This is where executable MATLAB scripts are
	stored which reproduce the results found in the
	article.

2) MSClassifyLib

	This folder contains parts of a larger library 
	developed at the Center for Industrial Mathematics,
	Bremen. It provides routines for the analysis of mass 
	spectrometry data and is required for the execution 
	of the scripts in the CNNs_Bioinformatics folder.

###########################################################

In-depth description of the files:

1) CNNs_Bioinformatics

	This folder contains two subfolders, one for the
	creation of the classification models and one for
	their analysis and evaluation.
	NOTE: In order to execute these scripts, the 
	MSClassifyLib needs to be included in MATLAB's
	search path.

	a) Training:
		The contained files compute the models
		for LDA with peak picking through their
		Mann-Whitney-Wilcoxon statistic as well
		as neural networks with the residual
		network architecture and IsotopeNet.

		NOTE: The file paths for DataTaskADSQ.mat
		and DataTaskLP.mat need to be adjusted.

	b) Evaluation:
		analyzeCNNmodel.m creates various statistics
		of the computed classification models.
		evaluationBoxplot.m compares the various
		models using box plots.
61 62 63 64
		
	c) IsotopeSizeEstimation:
	    scripts for the estimation of isotope sizes 
	    (see supplementatl material, section 1.1.1)
Jens Behrmann's avatar
Jens Behrmann committed
65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99

2) MSClassifyLib

	This folder contains several subfolders, which
	comprise the MSClassifyLib.

	a) Core
		This folder contains most of the required
		definitions of the employed data types.

	b) Classification
		The employed classifiers are found here,
		including the definitions of the used
		neural network architectures.

	c) ClassificationValidation
		This folder contains classes and methods
		with which a classification pipeline may
		be defined.

	d) FeatureExtraction
		This folder contains classes and methods
		for feature extraction, in particular
		for computing the Mann-Whitney-Wilcoxon 
		statistic. Furthermore, 'dummy' feature
		extraction methods are defined, which are
		required for the classification pipeline.

	e) Preprocessing
		Several preprocessing methods. Here, only
		the ones needed for the classification 
		pipeline are provided.

	f) Helpers
		Some small additional classes and methods.