Clustering with ipython

In [52]:

    

#This line is very important: (It turns on inline the visuals!)
%pylab inline
import csv

#You will need these also. These functions extract the data from the results file.
def load_file_return_data(filepath):
    data =[]
    with open(filepath,'r') as f:
        reader=csv.reader(f,delimiter='\t')
        headers = reader.next()
        for line in reader:
            data.append(line)

    name_list = list(enumerate(headers))
    return data, name_list
def return_data_with_header(header,data,name_list):
    for idx, name in name_list:
        if name == header:
            ind_to_take = idx
    data_col = []
    for line in data:
        
        data_col.append(float(line[ind_to_take]))
    return np.array(data_col)


    

    




Populating the interactive namespace from numpy and matplotlib

In [53]:

    

#You insert the local path where your exported imageJ 
#where 'Results.txt' is currently written.

data,name_list = load_file_return_data('Results.txt')
print name_list
#Insert the header you wish to extract here:
header1 = 'Area'
data_col1 = return_data_with_header(header1,data,name_list)
#Insert the header you wish to extract here:
header2 = 'Mean'
data_col2 = return_data_with_header(header2,data,name_list)


    

    




[(0, ' '), (1, 'Area'), (2, 'Mean'), (3, 'StdDev'), (4, 'Mode'), (5, 'Min'), (6, 'Max'), (7, 'IntDen'), (8, 'Median'), (9, 'RawIntDen')]

In [106]:

    

plot(data_col1,data_col2, 'o')
title('Area Versuses mean intensity plot')
xlabel('Area')
ylabel('Mean intensity')


    

    
Out[106]:





<matplotlib.text.Text at 0x11899f150>






    

In [101]:

    

#To cluster the data we start by using the kmeans algorithm.
from sklearn.cluster import KMeans,SpectralClustering 
#We initialise a kmeans model in the variable kmeans:
kmeans = KMeans(n_clusters=2, init='k-means++', n_init=10, max_iter=300, tol=0.0001, precompute_distances='auto')
#For more information.
#http://scikit-learn.org/stable/modules/generated/sklearn.cluster.KMeans.html


    

In [102]:

    

#Now we reorganise the data to a format which is compatible with kmeans algorithm.
data_arr = np.zeros((data_col2.__len__(),2))
data_arr[:,0] = np.array(data_col1)
data_arr[:,1] = np.array(data_col2)
#Now we use this data to fit the kmeans model in an unsupervised fashion.
kmeans.fit(data_arr)
out = kmeans.predict(data_arr)


    

In [105]:

    

#Now we fit the two clusters we have tried to find
plot(data_col1[out == 1],data_col2[out ==1], 'go')
plot(data_col1[out == 0],data_col2[out ==0], 'ro')

title('Area Versuses mean intensity clustering plot')
xlabel('Area')
ylabel('Mean intensity')
#The clustering algorithm should have highlighted good proportion
#of the cells which are both large and green. These represent the
#Dendritic cells in the image.


    

    
Out[105]:





<matplotlib.text.Text at 0x1188e2f50>






    

In [ ]: