def test_convolution_specialized(): stimuli=np.array([0,5,15]) on_off1=np.array([0,1,0]) on_off2=np.array([1,0,1]) x=np.linspace(0,45,91) # 0, .5, 1, 1.5, 2, ... 45 HRF1=convolution_specialized(stimuli,on_off1,hrf_single,x) y1=np.array([hrf_single(x_i-5) for x_i in x]) # what it should be doing HRF2=convolution_specialized(stimuli,on_off2,hrf_single,x) y2=np.array([hrf_single(x_i)+hrf_single(x_i-15) for x_i in x]) #what it should be doing assert all(HRF1 == y1) assert all(HRF2 == y2)
for i,name in enumerate(cond_string): nueral_prediction = events2neural(condition_location+name,TR,n_vols) hrf_long = np.convolve(nueral_prediction, hrf_at_trs) X_np[:,i+1] = hrf_long[:-(len(hrf_at_trs)-1)] all_tr_times = np.arange(n_vols) * TR ############################### # convolution_specialized # ############################### X_my = np.ones((n_vols,4)) conds = [cond1[:,0],cond2[:,0],cond3[:,0]] for i,cond in enumerate(conds): X_my[:,i+1]=convolution_specialized(cond,np.ones(len(cond)),hrf_single,all_tr_times) ########## # GLM # ########## ################### # np.convolve # ################### B_np,junk=glm_multiple(data,X_np) ############################### # convolution_specialized #
testconv_np=testconv_np[:N] ##################### # b. user functions # ##################### #--------# # second # testconv_2 = convolution(all_tr_times,neural_prediction,hrf_single) #-------# # third # testconv_3 = convolution_specialized(all_tr_times,neural_prediction, hrf_single,all_tr_times) #--------# # fourth # on_off = np.zeros(174) real_times,on_off[:-1] = np.linspace(0,432.5,173+1),neural_prediction hrf_function,TR,record_cuts= hrf_single, 2.5 ,np.linspace(0,432.5,173+1) # testconv_4_1 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=1) testconv_4_15 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=15) testconv_4_30 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=30)
img = nib.load(pathtodata + "BOLD/task001_run001/bold.nii.gz")
data = img.get_data()
data = data[..., 6:] # Knock off the first 6 observations.
cond1 = np.loadtxt(condition_location + "cond001.txt")
cond2 = np.loadtxt(condition_location + "cond002.txt")
cond3 = np.loadtxt(condition_location + "cond003.txt")
#######################
# Smart convolution #
#######################
all_stimuli = np.array(
sorted(list(cond2[:, 0]) + list(cond3[:, 0]) +
list(cond1[:, 0]))) # could also just x_s_array
my_hrf = convolution_specialized(all_stimuli, np.ones(len(all_stimuli)),
hrf_single, np.linspace(0, 239 * 2 - 2, 239))
##################
# np.convolve #
##################
# Suppose that TR=2. We know this is not a good assumption.
# Also need to look into the hrf function.
# initial needed values
TR = 2
tr_times = np.arange(0, 30, TR)
hrf_at_trs = np.array([hrf_single(x) for x in tr_times])
n_vols = data.shape[-1]
# creating the .txt file for the events2neural function
cond_all = np.row_stack((cond1, cond2, cond3))
# Load the image data for subject 1. img = nib.load(pathtodata+"BOLD/task001_run001/bold.nii.gz") data = img.get_data() data = data[...,6:] # Knock off the first 6 observations. cond1=np.loadtxt(condition_location+"cond001.txt") cond2=np.loadtxt(condition_location+"cond002.txt") cond3=np.loadtxt(condition_location+"cond003.txt") ####################### # Smart convolution # ####################### all_stimuli=np.array(sorted(list(cond2[:,0])+list(cond3[:,0])+list(cond1[:,0]))) # could also just x_s_array my_hrf = convolution_specialized(all_stimuli,np.ones(len(all_stimuli)),hrf_single,np.linspace(0,239*2-2,239)) ################## # np.convolve # ################## # Suppose that TR=2. We know this is not a good assumption. # Also need to look into the hrf function. # initial needed values TR = 2 tr_times = np.arange(0, 30, TR) hrf_at_trs = np.array([hrf_single(x) for x in tr_times]) n_vols=data.shape[-1] # creating the .txt file for the events2neural function
all_tr_times = np.arange(n_vols) * TR
cond_all = np.loadtxt(condition_location + "cond_all.txt")
#################
# First Attempt #
#################
# First approach allowed for fourier strength to be fit to each voxel,
# potentially overcorrecting and masking some response to neural stimulation
# X matrix
X = np.ones((n_vols, 9)) # changed since fourier needs more
X[:, 1] = convolution_specialized(cond_all[:, 0], np.ones(len(cond_all)), hrf_single, all_tr_times)
X[:, 2] = np.linspace(-1, 1, num=X.shape[0]) # drift
X[:, 3:] = fourier_creation(X.shape[0], 3)[:, 1:]
# modeling voxel hemodynamic response
beta, junk = glm_multiple(data, X)
MRSS, fitted, residuals = glm_diagnostics(beta, X, data)
# individual voxel analysis
plt.plot(all_tr_times, data[41, 47, 2], label="actual", color="b")
plt.plot(all_tr_times, fitted[41, 47, 2], label="predicted", color="r")
plt.title("Data for sub001, voxel [41, 47, 2],fourier 3 fit to voxel")
plt.xlabel("Time")
plt.ylabel("Hemodynamic response")
plt.legend(loc="upper right", shadow=True, fontsize="smaller")
M = len(hrf_at_trs) # M == 12
testconv_np = testconv_np[:N]
#####################
# b. user functions #
#####################
#--------#
# second #
testconv_2 = convolution(all_tr_times, neural_prediction, hrf_single)
#-------#
# third #
testconv_3 = convolution_specialized(all_tr_times, neural_prediction,
hrf_single, all_tr_times)
#--------#
# fourth #
on_off = np.zeros(174)
real_times, on_off[:-1] = np.linspace(0, 432.5, 173 + 1), neural_prediction
hrf_function, TR, record_cuts = hrf_single, 2.5, np.linspace(0, 432.5, 173 + 1)
#
testconv_4_1 = np_convolve_30_cuts(real_times,
on_off,
hrf_function,
TR,
record_cuts,
cuts=1)
def test_convolution():
#################
# i. Can the user-created functions match np.convolve in np.convolve territory
TR = 2.5
tr_times = np.arange(0, 30, TR)
hrf_at_trs = np.array([hrf_single(x) for x in tr_times])
n_vols = 173
neural_prediction = events2neural(location_to_class_data+'ds114_sub009_t2r1_cond.txt',TR,n_vols)
all_tr_times = np.arange(173) * TR
##################
# a. np.convolve #
##################
testconv_np = np.convolve(neural_prediction, hrf_at_trs) # hrf_at_trs sample data
N = len(neural_prediction) # N == n_vols == 173
M = len(hrf_at_trs) # M == 12
testconv_np=testconv_np[:N]
#####################
# b. user functions #
#####################
#--------#
# second #
testconv_2 = convolution(all_tr_times,neural_prediction,hrf_single)
#-------#
# third #
testconv_3 = convolution_specialized(all_tr_times,neural_prediction,
hrf_single,all_tr_times)
#--------#
# fourth #
on_off = np.zeros(174)
real_times,on_off[:-1] = np.linspace(0,432.5,173+1),neural_prediction
hrf_function,TR,record_cuts= hrf_single, 2.5 ,np.linspace(0,432.5,173+1)
#
testconv_4_1 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=1)
testconv_4_15 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=15)
testconv_4_30 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=30)
#-------#
# fifth #
testconv_5 = fast_convolution(all_tr_times,neural_prediction,fast_hrf,all_tr_times)
additional_runs=[testconv_np,testconv_2,testconv_3,testconv_4_1,testconv_4_15,testconv_4_30,testconv_5]
names=["testconv_np","testconv_2","testconv_3","testconv_4_1","testconv_4_15","testconv_4_30","testconv_5"]
print("Max difference between model and testconv_np:")
for i,my_convolved in enumerate(additional_runs):
if my_convolved.shape[0]==testconv_np.shape[0]:
print(names[i],max(abs(testconv_np-my_convolved)))
else:
print(names[i],max(abs(testconv_np-my_convolved[:-1])))
# Actual asserts
for i,my_convolved in enumerate(additional_runs):
if my_convolved.shape[0]==testconv_np.shape[0]:
assert (max(abs(testconv_np-my_convolved) < .0001))
else:
assert (max(abs(testconv_np-my_convolved[:-1]) < .0001))
for i,name in enumerate(cond_string): nueral_prediction = events2neural(condition_location+name,TR,n_vols) hrf_long = np.convolve(nueral_prediction, hrf_at_trs) X_np[:,i+1] = hrf_long[:-(len(hrf_at_trs)-1)] all_tr_times = np.arange(n_vols) * TR ############################### # convolution_specialized # ############################### X_my = np.ones((n_vols,4)) conds = [cond1[:,0],cond2[:,0],cond3[:,0]] for i,cond in enumerate(conds): X_my[:,i+1]=convolution_specialized(cond,np.ones(len(cond)),hrf_single,all_tr_times) ########## # GLM # ########## ################### # np.convolve # ################### B_np,junk=glm_multiple(data,X_np) ############################### # convolution_specialized #
one_zeros[16:20]=1
plt.scatter(np.arange(40),one_zeros)
plt.xlim(-1,40)
plt.title("Stimulus pattern")
plt.savefig(location_of_created_images+"on_off_pattern.png")
plt.close()
plt.plot(np.linspace(0,30,200),np.array([hrf_single(x) for x in np.linspace(0,30,200)]))
plt.title("Single HRF, started at t=0")
plt.savefig(location_of_created_images+"hrf_pattern.png")
plt.close()
convolved=convolution_specialized(np.arange(40),one_zeros,hrf_single,np.linspace(0,60,300))
plt.plot(np.linspace(0,60,300),convolved)
plt.title("Convolution")
plt.savefig(location_of_created_images+"initial_convolved.png")
plt.close()
colors=["#CCCCFF","#C4C3D0","#92A1CF","#2A52BE","#003399","#120A8F","#000080","#002366"]
one_zeros[4] = 1
one_zeros[16:20] = 1
plt.scatter(np.arange(40), one_zeros)
plt.xlim(-1, 40)
plt.title("Stimulus pattern")
plt.savefig(location_of_created_images + "on_off_pattern.png")
plt.close()
plt.plot(np.linspace(0, 30, 200),
np.array([hrf_single(x) for x in np.linspace(0, 30, 200)]))
plt.title("Single HRF, started at t=0")
plt.savefig(location_of_created_images + "hrf_pattern.png")
plt.close()
convolved = convolution_specialized(np.arange(40), one_zeros, hrf_single,
np.linspace(0, 60, 300))
plt.plot(np.linspace(0, 60, 300), convolved)
plt.title("Convolution")
plt.savefig(location_of_created_images + "initial_convolved.png")
plt.close()
colors = [
"#CCCCFF", "#C4C3D0", "#92A1CF", "#2A52BE", "#003399", "#120A8F",
"#000080", "#002366"
]
xx = np.linspace(0, 30, 3001)
i = 3
one_zeros_2 = np.zeros(3001)
one_zeros_2[(2 * i * 100 - 15):(2 * i * 100 + 15)] = .6
TR = 2
all_tr_times = np.arange(n_vols) * TR
cond_all = np.loadtxt(condition_location + "cond_all.txt")
#################
# First Attempt #
#################
# First approach allowed for fourier strength to be fit to each voxel,
# potentially overcorrecting and masking some response to neural stimulation
# X matrix
X = np.ones((n_vols, 9)) #changed since fourier needs more
X[:, 1] = convolution_specialized(cond_all[:, 0], np.ones(len(cond_all)),
hrf_single, all_tr_times)
X[:, 2] = np.linspace(-1, 1, num=X.shape[0]) #drift
X[:, 3:] = fourier_creation(X.shape[0], 3)[:, 1:]
# modeling voxel hemodynamic response
beta, junk = glm_multiple(data, X)
MRSS, fitted, residuals = glm_diagnostics(beta, X, data)
# individual voxel analysis
plt.plot(all_tr_times, data[41, 47, 2], label="actual", color="b")
plt.plot(all_tr_times, fitted[41, 47, 2], label="predicted", color="r")
plt.title("Data for sub001, voxel [41, 47, 2],fourier 3 fit to voxel")
plt.xlabel("Time")
plt.ylabel("Hemodynamic response")
plt.legend(loc='upper right', shadow=True, fontsize="smaller")
cond_all = np.array(
sorted(list(cond2[:, 0]) + list(cond3[:, 0]) +
list(cond1[:, 0]))) # could also just x_s_array
#--------#
# second #
#--------#
conv_2 = convolution(cond_all, np.ones(len(cond_all)), hrf_single)
scaled_2 = (conv_2 - np.mean(conv_2)) / (2 * np.std(conv_2)) + .4
#-------#
# third #
#-------#
conv_3 = convolution_specialized(cond_all, np.ones(len(cond_all)), hrf_single,
np.linspace(0, 239 * 2 - 2, 239))
scaled_3 = (conv_3 - np.mean(conv_3)) / (2 * np.std(conv_3)) + .4
#--------#
# fourth #
#--------#
real_times, on_off = cond_all, np.ones(len(cond_all))
hrf_function, TR, record_cuts = hrf_single, 2, np.linspace(0, 239 * 2 - 2, 239)
conv_4_15 = np_convolve_30_cuts(real_times,
on_off,
hrf_function,
TR,
record_cuts,
cuts=15)
cond_all=np.array(sorted(list(cond2[:,0])+list(cond3[:,0])+list(cond1[:,0]))) # could also just x_s_array #--------# # second # #--------# conv_2 = convolution(cond_all,np.ones(len(cond_all)),hrf_single) scaled_2=(conv_2-np.mean(conv_2))/(2*np.std(conv_2)) +.4 #-------# # third # #-------# conv_3 = convolution_specialized(cond_all,np.ones(len(cond_all)),hrf_single,np.linspace(0,239*2-2,239)) scaled_3=(conv_3-np.mean(conv_3))/(2*np.std(conv_3)) +.4 #--------# # fourth # #--------# real_times,on_off = cond_all,np.ones(len(cond_all)) hrf_function,TR,record_cuts= hrf_single, 2 ,np.linspace(0,239*2-2,239) conv_4_15 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=15) scaled_4_15=(conv_4_15-np.mean(conv_4_15))/(2*np.std(conv_4_15)) +.4 conv_4_30 = np_convolve_30_cuts(real_times,on_off,hrf_function,TR,record_cuts,cuts=30)
