if the testing data was taking from small number of events then we can find the segments by measuring the distance between first samples of one segment and last samples of other segment and find the best assignment. We can then combine the scores given to each segment and even use overlap segments.
In [1]:
%matplotlib inline
from matplotlib import pylab as pl
import cPickle as pickle
import pandas as pd
import numpy as np
import os
import random
In [2]:
import sys
sys.path.append('..')
Extarct the first and last W data samples of each test segment
In [3]:
W = 10
In [83]:
data_type = 'test' # preictal interictal, test
In [126]:
import scipy.io
edges = []
for target in ['Patient_2']: #'Dog_1', 'Dog_2', 'Dog_3', 'Dog_4', 'Dog_5', 'Patient_1', 'Patient_2']:
for segment in range(1,1000000):
fname = '../seizure-data/%s/%s_%s_segment_%04d.mat'%(target,target,data_type,segment)
try:
data = scipy.io.loadmat(fname)
k = '%s_segment_%d'%(data_type,segment)
d = data[k]['data'][0,0]
edges.append((d[:,:W].astype(float),d[:,-W:].astype(float)))
except:
break
find the closest pair of segments in the test data
In [127]:
def mydist(j1, j2):
return -2.*np.dot(j1,j2)/(np.dot(j1,j1) + np.dot(j2,j2))
# return -np.dot(j1,j2)/np.sqrt(np.dot(j1,j1)*np.dot(j2,j2))
In [128]:
N = len(edges)
d = np.zeros((N,N))
for i in range(N):
ei = edges[i]
for j in range(i+1,N):
ej = edges[j]
d[i,j] = mydist(ei[1][:,-1],ej[0][:,0])
d[j,i] = mydist(ej[1][:,-1],ei[0][:,0])
In [129]:
for i in range(N):
d[i,i] = 1
In [130]:
np.save('../data-cache/%s-test-jump-distance'%target, d)
In [131]:
ord = np.unravel_index(d.ravel().argsort(),d.shape)
In [132]:
N/6
Out[132]:
In [133]:
pl.hist(d[ord],bins=50);
In [134]:
m = 0
s1,s2=np.array(ord).T[m]
s1,s2,d[s1,s2]
Out[134]:
In [135]:
fuse = np.concatenate((edges[s1][1][:,:], edges[s2][0][:,:]),axis=-1)
In [136]:
fuse = fuse.squeeze()
fuse.shape
Out[136]:
In [137]:
Nbase = 0
Nchannels = fuse.shape[0]
for i in range(Nchannels):
pl.subplot(Nchannels,1,i+1)
channel = i+Nbase
pl.plot(fuse[channel,:])
pl.gcf().set_size_inches(14,16);
In [138]:
N/6
Out[138]:
In [139]:
next_segment = [-1]*N
previous_segment = [-1]*N
for i,(s1,s2) in enumerate(np.array(ord).T):
dist = d[s1,s2]
if next_segment[s1] != -1:
print i,'right conflict',dist
break
if previous_segment[s2] != -1:
print i,'left conflict',dist
break
next_segment[s1] = s2
previous_segment[s2] = s1
In [46]:
s1,s2
Out[46]:
In [47]:
s22 = next_segment[s1]
s1,s22
Out[47]:
In [48]:
from scipy import signal
def plfuse(s1,s2,channel=0):
fuse = np.concatenate((edges[s1][1][:,:], edges[s2][0][:,:]),axis=-1).squeeze()
pl.subplot(3,1,1)
f, Pxx_den = signal.periodogram(edges[s1][1][channel,:])
pl.plot(f, Pxx_den);
pl.subplot(3,1,2)
f, Pxx_den = signal.periodogram(edges[s2][0][channel,:])
pl.plot(f, Pxx_den);
f, Pxx_den = signal.periodogram(fuse[channel,:])
pl.subplot(3,1,3)
pl.plot(f, Pxx_den);
In [49]:
plfuse(s1,s2)
In [50]:
plfuse(s1,s22)
In [ ]: