In [1]:

    

%matplotlib inline
import pandas as pd
from pandas.tools import plotting
import mia


    

In [2]:

    

raw = pd.DataFrame.from_csv('../results/2015-03-15-results-blobs.csv')
raw.index = raw.image_name
raw.head()


    

    
Out[2]:






  

    

      

      
x
      
y
      
radius
      
image_name
      
patient_id
      
view
      
side
      
class
    
    

      
image_name
      

      

      

      

      

      

      

      

    
  
  

    

      
p214-010-60001-cl.png
      
 1639
      
 536
      
 128.000000
      
 p214-010-60001-cl.png
      
 21401060001
      
 c
      
 l
      
 3
    
    

      
p214-010-60001-cl.png
      
 1335
      
 328
      
 128.000000
      
 p214-010-60001-cl.png
      
 21401060001
      
 c
      
 l
      
 3
    
    

      
p214-010-60001-cl.png
      
 1867
      
 636
      
  64.000000
      
 p214-010-60001-cl.png
      
 21401060001
      
 c
      
 l
      
 3
    
    

      
p214-010-60001-cl.png
      
 2072
      
 648
      
  45.254834
      
 p214-010-60001-cl.png
      
 21401060001
      
 c
      
 l
      
 3
    
    

      
p214-010-60001-cl.png
      
 2016
      
 839
      
  45.254834
      
 p214-010-60001-cl.png
      
 21401060001
      
 c
      
 l
      
 3
    
  

In [3]:

    

features = mia.reduction.feature_statistics(raw)
features.describe()


    

    
Out[3]:






  

    

      

      
blob_count
      
avg_radius
      
std_radius
      
min_radius
      
max_radius
      
small_radius_count
      
med_radius_count
      
large_radius_count
      
density
      
lower_radius_qt
      
upper_radius_qt
      
upper_dist_count
      
patient_id
      
class
    
  
  

    

      
count
      
 360.000000
      
 360.000000
      
 360.000000
      
 360
      
 360.000000
      
 360.000000
      
 360.000000
      
 360.000000
      
 360.000000
      
 360.000000
      
 360.000000
      
 360.000000
      
 3.600000e+02
      
 360.000000
    
    

      
mean
      
 156.852778
      
  18.693714
      
  21.523458
      
   8
      
 153.189790
      
 151.250000
      
   2.725000
      
   2.877778
      
  60.871859
      
   8.197902
      
  18.691750
      
  41.675000
      
 2.140106e+10
      
   2.533333
    
    

      
std
      
  99.265677
      
   4.080252
      
   7.966148
      
   0
      
  39.602474
      
  98.051511
      
   3.780088
      
   1.566201
      
  13.039158
      
   0.764154
      
   4.394923
      
  32.620762
      
 5.295273e+02
      
   0.958178
    
    

      
min
      
  26.000000
      
  12.256703
      
   6.094591
      
   8
      
  45.254834
      
  23.000000
      
   0.000000
      
   1.000000
      
  42.650978
      
   8.000000
      
  11.313708
      
   5.000000
      
 2.140106e+10
      
   1.000000
    
    

      
25%
      
  82.000000
      
  15.775176
      
  15.420732
      
   8
      
 128.000000
      
  77.750000
      
   1.000000
      
   2.000000
      
  51.612142
      
   8.000000
      
  16.000000
      
  20.000000
      
 2.140106e+10
      
   2.000000
    
    

      
50%
      
 132.500000
      
  17.922227
      
  21.204195
      
   8
      
 181.019336
      
 127.500000
      
   2.000000
      
   2.000000
      
  57.892584
      
   8.000000
      
  16.000000
      
  31.000000
      
 2.140106e+10
      
   3.000000
    
    

      
75%
      
 200.250000
      
  20.493649
      
  26.711628
      
   8
      
 181.019336
      
 190.500000
      
   3.000000
      
   4.000000
      
  66.955485
      
   8.000000
      
  22.627417
      
  52.000000
      
 2.140106e+10
      
   3.000000
    
    

      
max
      
 631.000000
      
  37.188077
      
  51.391117
      
   8
      
 181.019336
      
 628.000000
      
  34.000000
      
   8.000000
      
 111.438717
      
  11.313708
      
  45.254834
      
 241.000000
      
 2.140106e+10
      
   4.000000
    
  

In [4]:

    

r = raw[raw['radius'] > 30]
mia.plotting.plot_risk_classes(r, 'radius')


    

In [16]:

    

mia.plotting.plot_risk_classes(features, 'upper_dist_count')


    

In [30]:

    

selected_columns = ['blob_count', 'avg_radius', 'std_radius', 'upper_radius_qt']
selected_features = features[selected_columns]
selected_features.describe()


    

    
Out[30]:






  

    

      

      
blob_count
      
avg_radius
      
std_radius
      
upper_radius_qt
    
  
  

    

      
count
      
 360.000000
      
 360.000000
      
 360.000000
      
 360.000000
    
    

      
mean
      
 156.852778
      
  18.693714
      
  21.523458
      
  18.691750
    
    

      
std
      
  99.265677
      
   4.080252
      
   7.966148
      
   4.394923
    
    

      
min
      
  26.000000
      
  12.256703
      
   6.094591
      
  11.313708
    
    

      
25%
      
  82.000000
      
  15.775176
      
  15.420732
      
  16.000000
    
    

      
50%
      
 132.500000
      
  17.922227
      
  21.204195
      
  16.000000
    
    

      
75%
      
 200.250000
      
  20.493649
      
  26.711628
      
  22.627417
    
    

      
max
      
 631.000000
      
  37.188077
      
  51.391117
      
  45.254834
    
  

In [37]:

    

%matplotlib qt
mapping = mia.analysis.tSNE(selected_features, n_components=3, learning_rate=200, perplexity=35,
                            early_exaggeration=4.0, verbose=2)
mapping['class'] = features['class']
# mia.plotting.plot_scatter_2d(mapping, [0,1], labels='class')
mia.plotting.plot_scatter_3d(mapping, [0,1,2], labels=mapping['class'])


    

    




[t-SNE] Computing pairwise distances...
[t-SNE] Computed conditional probabilities for sample 360 / 360
[t-SNE] Mean sigma: 0.488665
[t-SNE] Iteration 10: error = 16.3209470, gradient norm = 0.1531477
[t-SNE] Iteration 20: error = 12.1055069, gradient norm = 0.1728686
[t-SNE] Iteration 30: error = 11.4114051, gradient norm = 0.1705614
[t-SNE] Iteration 40: error = 11.6730105, gradient norm = 0.1540901
[t-SNE] Iteration 50: error = 11.7546230, gradient norm = 0.1419830
[t-SNE] Iteration 60: error = 12.2627918, gradient norm = 0.1569128
[t-SNE] Iteration 70: error = 11.2874802, gradient norm = 0.1600637
[t-SNE] Iteration 80: error = 11.4941722, gradient norm = 0.1550822
[t-SNE] Iteration 83: did not make any progress during the last 30 episodes. Finished.
[t-SNE] Error after 83 iterations with early exaggeration: 11.176040
[t-SNE] Iteration 90: error = 0.5433609, gradient norm = 0.0200394
[t-SNE] Iteration 100: error = 0.3091525, gradient norm = 0.0080954
[t-SNE] Iteration 110: error = 0.2608823, gradient norm = 0.0041306
[t-SNE] Iteration 120: error = 0.2478559, gradient norm = 0.0025223
[t-SNE] Iteration 130: error = 0.2406393, gradient norm = 0.0024766
[t-SNE] Iteration 140: error = 0.2368743, gradient norm = 0.0010233
[t-SNE] Iteration 150: error = 0.2378585, gradient norm = 0.0007322
[t-SNE] Iteration 160: error = 0.2381354, gradient norm = 0.0006945
[t-SNE] Iteration 170: did not make any progress during the last 30 episodes. Finished.
[t-SNE] Error after 170 iterations: 0.238122