Adiopocyte analysis

Adipocytes are adipose tissue cells and are the body's main energy store. Their main role is to store energy in the form of triglycerides. In a situation of higher energy demand, the body can derive the substrates it needs from the stored reserves through the process of lipolysis, i.e. breakdown. On the other hand, when the food supply is too high and the amount of energy expended by the body is lower, lipogenesis will occur and the excess will be deposited in the fat cells.

Table of Contents

1. Adiopocyte analysis
   • Data
   • Aim
   • Steps
   • Results
2. Bibliography

Data

Input: H&E 40x digital images in .tif format.

Aim

1. Analysis of fat cells morphology (area, shape and distance from other cells) in the bedding of the small intestine.
2. To see if correlations can be found between these characteristics and variables such as gender or the need for reccurent surgery.

Steps

Notebook fatcells_features.ipynb

Fat cell segmentation and feature extraction file program - authors of the original paper see: https://github.com/abebe9849/Crohn_wsi

Notebook cell_features_analysis.ipynb

Analysis of the data and matching it with patient information:

• search for correlations (Pearson)
• analysis of differences between groups (Ranksum)
• clustering (Clustergram)

Results

Correlation analysis

The first step was to look for correlations between fat cell features and patient characteristics. It was particularly important to check the correlation between the features and whether the patient required re-operation. The alpha used was equal 0.05. The obtained correlation values, which are statistically significant, have a value between |0.3| and |0.5|, so are rather at a weak/medium.

Table 1. Patient features vs Fat Cell Areas features

Patient feature	Fat Cell feature	Pearson corr	P value
AgeAtSurgery	mean area	0.4492	0.0017
AgeAtSurgery	median area	0.3455	0.0187
AgeAtSurgery	area variance	0.4279	0.0030
AgeAtSurgery	area std	0.4814	0.0007
AgeAtSurgery	area 3rd quartile	0.4632	0.0012
AgeAtSurgery	area quartile range	0.5038	0.0003

Table 2. Patient features vs Fat Cell Background features

Patient feature	Fat Cell feature	Pearson corr	P value
AgeAtDiagnosis	background mean	-0.3254	0.0311
AgeAtDiagnosis	background skewness	-0.3732	0.0386
AgeAtDiagnosis	background 3rd quartile	-0.3398	0.0240
AgeAtSurgery	background mean	-0.4091	0.0048
AgeAtSurgery	background median	-0.3630	0.0132
AgeAtSurgery	background std	-0.3602	0.0139
AgeAtSurgery	background skewness	-0.3832	0.0304
AgeAtSurgery	background range	-0.3902	0.0073
AgeAtSurgery	background 1st quartile	-0.3165	0.0321
AgeAtSurgery	background 3rd quartile	-0.4488	0.0018
AgeAtSurgery	background quartile range	-0.3560	0.0152
Any_rec_withOUT_i2ab_6mo	background kurtosis	-0.4681	0.0120
Any_rec_WITH_i2ab_6mo	background kurtosis	-0.4681	0.0120

Table 3. Patient features vs Fat Cell Distances features

Patient feature	Fat Cell feature	Pearson corr	P value
AgeAtDiagnosis	distanes skewness	-0.3759	0.0119
AgeAtSurgery	distanes skewness	-0.3544	0.0157
Early_rec_withOUTi2ab_15mo	distances variance	-0.3683	0.0381
Early_rec_withOUTi2ab_15mo	distances std	-0.3737	0.0351
Early_rec_withOUTi2ab_15mo	distances quartile range	-0.4157	0.0180
Early_rec_WITHi2ab_15mo	distanece variation	-0.3683	0.0381
Early_rec_WITHi2ab_15mo	distance std	-0.3737	0.0351
Early_rec_WITHi2ab_15mo	distances quartile range	-0.4157	0.0180

Table 4. Patient features vs Fat Cell Flatness features

Patient feature	Fat Cell feature	Pearson corr	P value
Early_rec_withOUTi2ab_15mo	flatness range	-0.4192	0.0169
Early_rec_WITHi2ab_15mo	flatness range	-0.4192	0.0169
Any_rec_withOUT_i2ab_6mo	flatness median	0.3260	0.0401
Any_rec_WITH_i2ab_6mo	flatness median	0.3260	0.0401

Wilcoxon rank-sum test

The Wilcoxon rank-sum tests the null hypothesis that two sets of measurements are drawn from the same distribution. In this case, characteristics that divided patients into distinct groups (0/1) were analysed.

No differences were noted for the groups: Sex and Reccurent Surgery. The differences for the other groups are shown below.

Table 5. Differences in Early rec withOUT

Fat Cell feature	Statistic	P value
Background 1st quartile	2.1273	0.0334
Background 3rd quartile	2.0513	0.0402
Distance quartile range	2.2032	0.0276
Flatness range	2.0893	0.0367

Table 6. Differences in Early rec WITH

Fat Cell feature	Statistic	P value
Background 1st quartile	2.1273	0.0334
Background 3rd quartile	2.0513	0.0403
Distance quartile range	2.2033	0.0276
Flatness range	2.0893	0.0367

Table 7. Differences in Any rec withOUT

Fat Cell feature	Statistic	P value
Background mean	2.1840	0.0290
Background median	2.0660	0.0388
Background 1st quartile	2.0365	0.0417
Background 3rd quartile	2.2136	0.0269

Table 8. Differences in Any rec WITH

Fat Cell feature	Statistic	P value
Background mean	2.1840	0.0289
Background median	2.0660	0.0388
Background 1st quartile	2.0365	0.0417
Background 3rd quartile	2.2136	0.0269

Clusterization results

For characteristics considered to be statistically differentiating for Early rec withOUT/WITH and Any rec withOUT/WITH, clustering is possible:

Bibliography

• https://github.com/abebe9849/Crohn_wsi
• https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.ranksums.html