def classify(record, data_dir, clf=None):
x = loader.load_data_from_file(record, data_dir)
x = normalizer.normalize_ecg(x)
x = extractor.features_for_row(x)
if clf is None:
clf = get_saved_model(x.shape[1:])
# as we have one sample at a time to predict, we should resample it into 2d array to classify
x = np.array(x).reshape(1, -1)
return categorizer.get_original_label(clf.predict(x)[0])
for item in misclassified:
print(item[2], 'is of class', item[0], 'but was classified as',
item[1])
print(classification_report(ytrue, ypred))
matrix = confusion_matrix(ytrue, ypred)
print(matrix)
for row in matrix:
amax = sum(row)
if amax > 0:
for i in range(len(row)):
row[i] = row[i] * 100.0 / amax
print(matrix)
print("Plotting misclassified")
base_dir = "../outputs/misclassified"
mkdir(base_dir)
misclassified = sorted(misclassified, key=lambda t: t[2])
for item in misclassified:
print("Saving to " + item[2])
row = loader.load_data_from_file(item[2])
[ts, fts, rpeaks, tts, thb, hrts,
hr] = ecg.ecg(signal=row, sampling_rate=loader.FREQUENCY, show=False)
__draw_to_file(
base_dir + "/" + item[2] + "_" + item[0] + "_" + item[1] + ".png",
tts, thb)
plt.subplot(total, 1, 5)
plt.ylabel("QRS DFT")
ff = fft(qrs)[:len(qrs) // 2]
plt.plot(ff, 'r')
plt.subplot(total, 1, 6)
plt.ylabel(clazz)
plt.plot(t)
plt.subplot(total, 1, 7)
plt.ylabel("T DFT")
ff = fft(t)[:len(t) // 2]
plt.plot(ff, 'r')
plt.show()
# Normal: A00001, A00002, A0003, A00006
plot_with_detected_peaks(loader.load_data_from_file("A07088"), "Normal")
plot_with_detected_peaks(loader.load_data_from_file("A00006"), "Normal")
plot_with_detected_peaks(loader.load_data_from_file("A08128"), "Normal")
# AF: A00004, A00009, A00015, A00027
plot_with_detected_peaks(loader.load_data_from_file("A00004"), "AF rhythm")
plot_with_detected_peaks(loader.load_data_from_file("A00009"), "AF rhythm")
# Other: A00005, A00008, A00013, A00017
plot_with_detected_peaks(loader.load_data_from_file("A00005"), "Other rhythm")
plot_with_detected_peaks(loader.load_data_from_file("A00008"), "Other rhythm")
# Noisy: A00205, A00585, A01006, A01070
plot_with_detected_peaks(loader.load_data_from_file("A00205"), "Noisy signal")
plot_with_detected_peaks(loader.load_data_from_file("A00585"), "Noisy signal")
row,
'k-',
[x / 300 for x in p],
[row[x] for x in p],
'go',
[x / 300 for x in q],
[row[x] for x in q],
'bv',
[x / 300 for x in r],
[row[x] for x in r],
'r^',
[x / 300 for x in s],
[row[x] for x in s],
'bv',
[x / 300 for x in t],
[row[x] for x in t],
'mo',
)
plt.title("Positions of P, Q, R, S, T")
plt.xlabel("Time, s")
plt.ylabel("Normalized ECG signal")
plt.show()
# Normal: A00001, A00002, A0003, A00006
# AF: A00004, A00009, A00015, A00027
# Other: A00005, A00008, A00013, A00017
# Noisy: A00205, A00585, A01006, A01070
plot_with_detected_peaks(loader.load_data_from_file("A00001"), "Normal")
import numpy as np
model_file = "../model.pkl"
name = input("Enter an entry name to plot: ")
name = name.strip()
if len(name) == 0:
print("Finishing")
exit(-1)
if not loader.check_has_example(name):
print("File Not Found")
exit(-1)
x = loader.load_data_from_file(name, "../validation")
x = normalizer.normalize_ecg(x)
feature_names = feature_extractor5.get_feature_names(x)
x = feature_extractor5.features_for_row(x)
# as we have one sample at a time to predict, we should resample it into 2d array to classify
x = np.array(x).reshape(1, -1)
model = joblib.load(model_file)
dtree = model.estimators_[0]
def export_tree(dtree, feature_names):
dot_data = sklearn.tree.export_graphviz(dtree,
feature_names=feature_names,
out_file=None)
plt.plot(range(len(d2)), d2, 'g-')
plt.ylabel("D2")
plt.subplot(total, 1, 5)
plt.plot(range(len(d3)), d3, 'g-')
plt.ylabel("D3")
plt.subplot(total, 1, 6)
plt.plot(range(len(d4)), d4, 'g-')
plt.ylabel("D4")
plt.show()
# Normal: A00001, A00002, A0003, A00006
plot_wavelet(loader.load_data_from_file("A00001"), "Normal")
# AF: A00004, A00009, A00015, A00027
plot_wavelet(loader.load_data_from_file("A00004"), "AF rhythm")
plot_wavelet(loader.load_data_from_file("A00009"), "AF rhythm")
plot_wavelet(loader.load_data_from_file("A00015"), "AF rhythm")
plot_wavelet(loader.load_data_from_file("A00027"), "AF rhythm")
# Other: A00005, A00008, A00013, A00017
plot_wavelet(loader.load_data_from_file("A00005"), "Other rhythm")
plot_wavelet(loader.load_data_from_file("A00008"), "Other rhythm")
plot_wavelet(loader.load_data_from_file("A00013"), "Other rhythm")
plot_wavelet(loader.load_data_from_file("A00017"), "Other rhythm")
# Noisy: A00205, A00585, A01006, A01070
plot_wavelet(loader.load_data_from_file("A00205"), "Noisy signal")
plot_wavelet(loader.load_data_from_file("A00585"), "Noisy signal")
plot_wavelet(loader.load_data_from_file("A01006"), "Noisy signal")
plot_wavelet(loader.load_data_from_file("A01070"), "Noisy signal")
# templates = ecg.ecg(x, sampling_rate=300, show=False)['templates']
fs, powers = signal.welch(x, loader.FREQUENCY)
fs = fs[:40]
powers = powers[:40]
print(hrv.frequency_domain(x, fs=300))
plt.plot(fs, powers)
plt.xticks([2 * i for i in range(len(fs) // 2)])
plt.grid()
plt.show()
plot_powers(loader.load_data_from_file("A00001"))
plot_powers(loader.load_data_from_file("A07718"))
plot_powers(loader.load_data_from_file("A08523"))
plot_powers(loader.load_data_from_file("A00004"))
plot_powers(loader.load_data_from_file("A06746"))
plot_powers(loader.load_data_from_file("A07707"))
plot_powers(loader.load_data_from_file("A00013"))
plot_powers(loader.load_data_from_file("A06245"))
plot_powers(loader.load_data_from_file("A07908"))
plot_powers(loader.load_data_from_file("A01006"))
plot_powers(loader.load_data_from_file("A08402"))
plot_powers(loader.load_data_from_file("A04853"))
row[i] = row[i] * 100.0 / amax
print(matrix)
while (True):
name = input("Enter an entry name to plot: ")
name = name.strip()
if len(name) == 0:
print("Finishing")
break
if not loader.check_has_example(name):
print("File Not Found")
continue
row = loader.load_data_from_file(name)
print(info[name])
[ts, fts, rpeaks, tts, thb, hrts,
hr] = ecg.ecg(signal=row, sampling_rate=loader.FREQUENCY, show=True)
b = heartbeats.median_heartbeat(thb)
t = b[int(0.3 * loader.FREQUENCY):int(0.55 * loader.FREQUENCY)]
a = normalizer.normalize_ecg(t)
a[a < 0] = 0
a = np.square(a)
a -= np.mean(a)
import matplotlib
matplotlib.use("Qt5Agg")
import matplotlib.pyplot as plt
from loading import loader
from features.qrs_detect import *
plt.rcParams["figure.figsize"] = (14, 5)
seed = 42
random.seed(seed)
np.random.seed(seed)
print("Seed =", seed)
a_ecg = loader.load_data_from_file("A01171")
n_ecg = loader.load_data_from_file("A03823")
o_ecg = loader.load_data_from_file("A07915")
s_ecg = loader.load_data_from_file("A04946")
total = 4
plt.subplot(total, 1, 1)
plt.ylabel("A-Fib")
plt.plot(a_ecg)
plt.subplot(total, 1, 2)
plt.ylabel("Normal")
plt.plot(n_ecg)
plt.subplot(total, 1, 3)
plt.ylabel("Other")
print('R', len(r), r)
times = np.diff(r)
print(np.mean(times), np.std(times))
plt.subplot(2, 1, 1)
plt.plot(range(len(row)), row, 'g-',
r, [row[x] for x in r], 'r^')
plt.ylabel(clazz + "QRS 1")
r = qrs_detect2(row, fs=300, thres=0.4, ref_period=0.2)
print('R', len(r), r)
times = np.diff(r)
print(np.mean(times), np.std(times))
plt.subplot(2, 1, 2)
plt.plot(range(len(row)), row, 'g-',
r, [row[x] for x in r], 'r^')
plt.ylabel(clazz + " QRS 2")
plt.show()
# Normal: A00001, A00002, A0003, A00006
plot_with_detected_peaks(loader.load_data_from_file("A00001"), "Normal")
# AF: A00004, A00009, A00015, A00027
plot_with_detected_peaks(loader.load_data_from_file("A00004"), "AF")
# Other: A00005, A00008, A00013, A00017
plot_with_detected_peaks(loader.load_data_from_file("A00005"), "Other")
# Noisy: A00205, A00585, A01006, A01070
plot_with_detected_peaks(loader.load_data_from_file("A00205"), "Noisy")
