def preprocess(index):
# load data
t, stim, rec = load_frame_group(path, index=index, stacked=False)
# apply filter
rec_filt = []
for rec_ in rec:
rec_ = butter_lpf(rec_, lo_cutoff, fs)
rec_filt.append(rec_)
# mean
rec = np.stack(rec, axis=0).mean(axis=0)
rec = subtract_baseline(t, rec)
# filter
rec_filt = np.stack(rec_filt, axis=0).mean(axis=0)
rec_filt = subtract_baseline(t, rec_filt)
# split data by stimuli
ts1 = np.argmax(stim)
# mask
stim[ts1:ts1 + 100] = 0
ts2 = np.argmax(stim)
ts1, ts2 = t[ts1], t[ts2]
logger.debug("stimuli timestamp: {}, {}".format(ts1, ts2))
# split stimuli
t_, rec1 = t_crop(t, rec, (ts1, ts2))
_, rec_filt1 = t_crop(t, rec_filt, (ts1, ts2))
_, rec2 = t_crop(t, rec, (ts2, 2 * ts2 - ts1))
_, rec_filt2 = t_crop(t, rec_filt, (ts2, 2 * ts2 - ts1))
# offset t
t_ -= t_[0]
return t_, [rec1, rec2], [rec_filt1, rec_filt2]
def preprocess(index, crop=(0.1, 0.15)):
"""
Preprocess data directly from file.
Args:
index (ndarray): Frame range to extract, must be contiguous.
crop (tuple): Timestamp range for coarse croping.
Returns:
(tuple): tuple containing:
t (ndarray): Cropped timestamp.
rec (ndarray): Raw recordings.
rec_filt (ndarray): Filtered recordings.
"""
# load data
t, stim, rec = load_frame_group(path, index=index, stacked=False)
t_ = None
rec_tmp, rec_filt = [], []
for rec_ in rec:
rec_lpf = butter_lpf(rec_, lo_cutoff, fs)
rec_lpf = subtract_baseline(t, rec_lpf)
t_, rec_lpf = t_crop(t, rec_lpf, crop)
rec_filt.append(rec_lpf)
rec_ori = subtract_baseline(t, rec_)
_, rec_ori = t_crop(t, rec_ori, crop)
rec_tmp.append(rec_ori)
rec = rec_tmp
# offset t
t_ -= t_[0]
return t_, rec, rec_filt
def preprocess(index):
# load data
t, stim, rec = load_frame_group(path, index=index, stacked=False)
# determine stimuli split point
ts1 = np.argmax(stim)
# mask first stimuli
stim[ts1:ts1 + 100] = 0
ts2 = np.argmax(stim)
ts1, ts2 = t[ts1], t[ts2]
logger.debug("stimuli timestamp: {}, {}".format(ts1, ts2))
logger.info("applying LPF and background subtraction")
# apply filter and subtract baseline
rec_tmp, rec_filt = [], []
for rec_ in rec:
rec_lpf = butter_lpf(rec_, lo_cutoff, fs)
rec_lpf = subtract_baseline(t, rec_lpf)
rec_filt.append(rec_lpf)
rec_ori = subtract_baseline(t, rec_)
rec_tmp.append(rec_ori)
rec = rec_tmp
logger.info("cropping")
# split stimuli
t_ = None
rec1, rec2 = [], []
rec_filt1, rec_filt2 = [], []
for rec_, rec_filt_ in zip(rec, rec_filt):
t_, rec1_ = t_crop(t, rec_, (ts1, ts2))
rec1.append(rec1_)
_, rec_filt1_ = t_crop(t, rec_filt_, (ts1, ts2))
rec_filt1.append(rec_filt1_)
_, rec2_ = t_crop(t, rec_, (ts2, 2 * ts2 - ts1))
rec2.append(rec2_)
_, rec_filt2_ = t_crop(t, rec_filt_, (ts2, 2 * ts2 - ts1))
rec_filt2.append(rec_filt2_)
# offset t
t_ -= t_[0]
return (
t_,
[np.stack(rec1, axis=0),
np.stack(rec2, axis=0)],
[np.stack(rec_filt1, axis=0),
np.stack(rec_filt2, axis=0)],
)
def main(index, name="filter_demo_sink"):
plt.cla()
# load data
t, stim, rec = load_frame_group(path, index=index, stacked=False)
# apply filter
rec_filt = []
for rec_ in rec:
rec_ = butter_lpf(rec_, lo_cutoff, fs)
rec_filt.append(rec_)
# mean
rec = np.stack(rec, axis=0).mean(axis=0)
rec = subtract_baseline(t, rec)
crop = (0.1, 0.15)
t_, rec = t_crop(t, rec, crop)
ax.plot(t_, rec, "r", label="raw", linewidth=1)
rec_filt = np.stack(rec_filt, axis=0).mean(axis=0)
rec_filt = subtract_baseline(t, rec_filt)
t_, rec_filt = t_crop(t, rec_filt, crop)
ax.plot(t_, rec_filt, "b", label="filtered", linewidth=1)
ax.axhline(0, color="k", linestyle=":", linewidth=0.5)
# labels
ax.legend()
plt.xlabel("Time (s)")
plt.ylabel("Intensity (mV)")
# final adjust
ax.set_xlim(crop)
plt.savefig("{}.png".format(name), dpi=300)
plt.waitforbuttonpress()
def preprocess(index, crop):
# load data
t, stim, rec = load_frame_group(path, index=index, stacked=False)
# apply filter
rec_filt = []
for rec_ in rec:
rec_ = butter_lpf(rec_, lo_cutoff, fs)
rec_filt.append(rec_)
# mean
rec = np.stack(rec, axis=0).mean(axis=0)
rec = subtract_baseline(t, rec)
# filter
rec_filt = np.stack(rec_filt, axis=0).mean(axis=0)
rec_filt = subtract_baseline(t, rec_filt)
# crop
t_, rec = t_crop(t, rec, crop)
_, rec_filt = t_crop(t, rec_filt, crop)
# offset t
t_ -= t_[0]
return t_, rec, rec_filt
def main(index, name="filter_demo_sink"):
plt.cla()
# load data
t, stim, rec = load_frame_group(path, index=index, stacked=False)
# apply filter
rec_filt = []
for rec_ in rec:
rec_ = butter_lpf(rec_, lo_cutoff, fs)
rec_filt.append(rec_)
# mean
rec = np.stack(rec, axis=0).mean(axis=0)
rec = subtract_baseline(t, rec)
# visualize raw data
crop = (0.1, 0.15)
t_, rec = t_crop(t, rec, crop)
ax.plot(t_, rec, "r", label="raw", linewidth=1)
ax.axhline(0, color="k", linestyle=":", linewidth=0.5)
# using filtered signal to extract slope
rec_filt = np.stack(rec_filt, axis=0).mean(axis=0)
rec_filt = subtract_baseline(t, rec_filt)
t_, rec_filt = t_crop(t, rec_filt, crop)
ipk, _ = find_epsp_peak(t_, rec_filt)
ax.scatter(
t_[ipk],
rec_filt[ipk],
marker="o",
edgecolors="b",
facecolor="none",
label="EPSP peak",
)
# slope
slope, r, dt, dy = epsp_slope(t_, rec, ipk, yf=rec_filt, return_pos=True)
ax.plot(dt, dy, "b:", label="EPSP slope", linewidth=1)
# labels
ax.legend()
plt.xlabel("Time (s)")
plt.ylabel("Intensity (mV)")
# create inset
axin = inset_axes(ax, width="30%", height="50%", loc=4, borderpad=3)
t_in, rec_in = t_crop(t_, rec, (0.1025, 0.1100))
axin.plot(t_in, rec_in, "r", label="raw", linewidth=1)
axin.scatter(
t_[ipk],
rec_filt[ipk],
marker="o",
edgecolors="b",
facecolor="none",
label="EPSP peak",
)
axin.plot(dt, dy, "b:", linewidth=2)
axin.set_xlim((0.1025, 0.1100))
# final adjust
ax.set_xlim(crop)
plt.savefig("{}_slope-{:.3f}_r-{:.3f}.png".format(name, slope, r), dpi=300)
plt.waitforbuttonpress()
