-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtsv_to_h5.py
284 lines (252 loc) · 12 KB
/
tsv_to_h5.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
將 measurements_auscultatory 與 measurements_oscillometric 兩個資料夾中的原始波型資料，
依據 sliding window 切分後，連同參與者、量測資訊與特徵資訊一起存成 h5 檔案。
重點修改：
1. 嘗試先以 header=None 讀取 tsv 檔案，若第一列資料無法轉為 float，則改用 header=0 讀取，
以避免把欄位名稱誤讀入導致全為 NaN。
2. 呼叫與 GUI app 相同的濾波函式（filter_ekg, filter_ppg, filter_tonometry），使得轉檔後訊號尺度一致。
3. 從 features.tsv 取得特徵資訊（baseline_sbp、baseline_dbp、delta_sbp、delta_dbp），並存入 h5 屬性。
4. 參與者性別直接以原始字串儲存（例如 "M" 或 "F"）。
"""
import os
import glob
import pandas as pd
import numpy as np
import h5py
from tqdm import tqdm
from scipy import signal
# ======================
# 參數設定
# ======================
QUALITY_THRESHOLD = 0.65
WINDOW_LENGTH = 4096 # 每個 sliding window 的點數
WINDOW_STEP = 64 # 視窗步幅
# 資料夾與檔案路徑（請依實際情況調整）
AUSCULTATORY_DIR = 'measurements_auscultatory'
OSCILLOMETRIC_DIR = 'measurements_oscillometric'
OUTPUT_DIR = 'output_h5_folder'
if not os.path.exists(OUTPUT_DIR):
os.makedirs(OUTPUT_DIR)
# Metadata 檔案
PARTICIPANTS_TSV = 'participants.tsv'
FEATURES_TSV = 'features.tsv' # 若無此檔案可忽略
AUSCULTATORY_META_TSV = 'measurements_auscultatory.tsv'
OSCILLOMETRIC_META_TSV = 'measurements_oscillometric.tsv'
# ======================
# 讀取 Metadata 檔案
# ======================
print("讀取 metadata 檔案...")
participants_df = pd.read_csv(PARTICIPANTS_TSV, sep='\t', dtype=str)
if os.path.exists(FEATURES_TSV):
features_df = pd.read_csv(FEATURES_TSV, sep='\t', dtype=str)
else:
features_df = None
ausc_meta_df = pd.read_csv(AUSCULTATORY_META_TSV, sep='\t', dtype=str)
osc_meta_df = pd.read_csv(OSCILLOMETRIC_META_TSV, sep='\t', dtype=str)
# ----------------------
# Helper 函式：根據 pid, phase, measurement id 取得量測 metadata
# ----------------------
def get_measurement_metadata(pid, phase, meas_id, method='auscultatory'):
meas_id_for_compare = meas_id.replace('_',' ')
df = ausc_meta_df if method == 'auscultatory' else osc_meta_df
meta_rows = df[(df['pid'] == pid) &
(df['phase'] == phase) &
(df['measurement'].str.replace("_", " ") == meas_id_for_compare)]
if not meta_rows.empty:
return meta_rows.iloc[0]
else:
return None
# ----------------------
# (選用) 濾波函式，與 GUI 內使用的相同
# ----------------------
def filter_ekg(x, fs):
sos_dc = signal.iirfilter(N=2, Wn=0.1/(fs/2), btype='highpass', ftype='butter', output='sos')
y = signal.sosfiltfilt(sos_dc, x)
wp = 40.0 / (fs/2.0)
ws = 45.0 / (fs/2.0)
sos_lp = signal.iirdesign(wp=wp, ws=ws, gpass=0.1, gstop=40, ftype='ellip', output='sos')
y = signal.sosfiltfilt(sos_lp, y)
w0 = 60.0 / (fs/2.0)
bw = w0 / 3.0
w1 = [w0 - bw/2.0, w0 + bw/2.0]
sos_notch = signal.iirfilter(N=6, rp=0.1, Wn=w1, btype='bandstop', ftype='cheby1', output='sos')
y = signal.sosfiltfilt(sos_notch, y)
return y
def filter_ppg(x, fs):
sos_hp = signal.butter(N=4, Wn=0.25/(fs/2.0), btype='highpass', output='sos')
y = signal.sosfiltfilt(sos_hp, x)
wp = 10.0 / (fs/2.0)
ws = 12.0 / (fs/2.0)
sos_lp = signal.iirdesign(wp=wp, ws=ws, gpass=1, gstop=60, ftype='ellip', output='sos')
y = signal.sosfiltfilt(sos_lp, y)
return y
def filter_tonometry(x, fs):
ws = 0.2/(fs/2.0)
wp = 0.3/(fs/2.0)
sos_hp = signal.iirdesign(wp=wp, ws=ws, gpass=1, gstop=60, ftype='ellip', output='sos')
y = signal.sosfiltfilt(sos_hp, x)
wp2 = 22.0 / (fs/2.0)
ws2 = 26.0 / (fs/2.0)
sos_lp = signal.iirdesign(wp=wp2, ws=ws2, gpass=0.1, gstop=40, ftype='ellip', output='sos')
y = signal.sosfiltfilt(sos_lp, y)
return y
# ----------------------
# 處理資料夾
# ----------------------
measurement_dirs = [
(AUSCULTATORY_DIR, 'auscultatory'),
(OSCILLOMETRIC_DIR, 'oscillometric')
]
for m_dir, method in measurement_dirs:
if not os.path.exists(m_dir):
print(f"資料夾 {m_dir} 不存在，跳過")
continue
subdirs = sorted([d for d in os.listdir(m_dir) if os.path.isdir(os.path.join(m_dir, d))])
for subdir in tqdm(subdirs, desc=f"處理 {m_dir} 子資料夾"):
subject_folder = os.path.join(m_dir, subdir)
default_pid = subdir
tsv_files = glob.glob(os.path.join(subject_folder, '*.tsv'))
for file in tsv_files:
filename = os.path.basename(file)
h5_filename = os.path.splitext(filename)[0] + '.h5'
h5_filepath = os.path.join(OUTPUT_DIR, h5_filename)
if os.path.exists(h5_filepath):
print(f"h5 檔案 {h5_filename} 已存在，跳過")
continue
parts = filename.split('.')
if parts[-1].lower() != 'tsv':
continue
if len(parts) == 3:
pid = default_pid
phase = parts[0]
meas_id = parts[1]
elif len(parts) >= 4:
pid = parts[0]
phase = parts[1]
meas_id = parts[2]
else:
print(f"檔名格式不符，略過: {filename}")
continue
measurement = meas_id
measurement_for_compare = measurement.replace("_", " ")
# -----------------------------
# 讀取 TSV 檔案
# -----------------------------
try:
data = pd.read_csv(file, sep='\t', header=None)
try:
float(data.iloc[0, 0])
except (ValueError, TypeError):
data = pd.read_csv(file, sep='\t', header=0)
except Exception as e:
print(f"讀取 {filename} 時發生錯誤: {e}，跳過")
continue
if data.shape[1] == 7:
ekg_signal = pd.to_numeric(data.iloc[:, 1], errors='coerce').values
ppg_signal = pd.to_numeric(data.iloc[:, 2], errors='coerce').values
tono_signal = pd.to_numeric(data.iloc[:, 3], errors='coerce').values
elif data.shape[1] == 6:
ekg_signal = pd.to_numeric(data.iloc[:, 0], errors='coerce').values
ppg_signal = pd.to_numeric(data.iloc[:, 1], errors='coerce').values
tono_signal = pd.to_numeric(data.iloc[:, 2], errors='coerce').values
else:
print(f"{filename} 欄位數 ({data.shape[1]}) 不符合預期，略過。")
continue
min_length = min(len(ekg_signal), len(ppg_signal), len(tono_signal))
if min_length < WINDOW_LENGTH:
print(f"{filename} 信號長度 ({min_length}) 少於視窗長度，略過。")
continue
ekg_signal = ekg_signal[:min_length]
ppg_signal = ppg_signal[:min_length]
tono_signal = tono_signal[:min_length]
# 估算 fs，若無時間資訊則假設等間隔
fs = None
if min_length > 1:
if data.shape[1] == 7:
x_data = pd.to_numeric(data.iloc[:, 0], errors='coerce').values
if np.all(~np.isnan(x_data)):
dt = np.mean(np.diff(x_data))
if dt > 0:
fs = 1.0 / dt
else:
fs = 500
if fs is not None:
ekg_signal = filter_ekg(ekg_signal, fs)
ppg_signal = filter_ppg(ppg_signal, fs)
tono_signal = filter_tonometry(tono_signal, fs)
ekg_windows = []
ppg_windows = []
tono_windows = []
for start in range(0, min_length - WINDOW_LENGTH + 1, WINDOW_STEP):
ekg_windows.append(ekg_signal[start:start+WINDOW_LENGTH])
ppg_windows.append(ppg_signal[start:start+WINDOW_LENGTH])
tono_windows.append(tono_signal[start:start+WINDOW_LENGTH])
if len(ekg_windows) == 0:
print(f"{filename} 無法切出視窗，略過。")
continue
ekg_windows = np.array(ekg_windows)
ppg_windows = np.array(ppg_windows)
tono_windows = np.array(tono_windows)
if np.isnan(ekg_windows[0]).all() or np.isnan(ppg_windows[0]).all() or np.isnan(tono_windows[0]).all():
print(f"警告：{filename} 第一個視窗有問題 (全 NaN)，請檢查檔案內容")
# -----------------------------
# 存成 h5 檔案
# -----------------------------
try:
with h5py.File(h5_filepath, 'w') as hf:
hf.create_dataset('EKG', data=ekg_windows)
hf.create_dataset('PPG', data=ppg_windows)
hf.create_dataset('Tonometry', data=tono_windows)
# 存入受試者資訊 (gender 直接存原始字串，如 "M" 或 "F")
part_row = participants_df[participants_df['pid'] == pid].iloc[0]
participant_info = {
"age": part_row.get("age", ""),
#gender編碼成0或1
"gender": 0 if part_row.get("gender", "") == "M" else 1,
"height": part_row.get("height", ""),
"weight": part_row.get("weight", "")
}
for key, value in participant_info.items():
hf.attrs[f"participant_{key}"] = value
# 存入量測資訊
current_meta_df = ausc_meta_df if method=='auscultatory' else osc_meta_df
rows = current_meta_df[
(current_meta_df['pid'] == pid) &
(current_meta_df['phase'] == phase) &
(current_meta_df['measurement'] == measurement_for_compare)
]
meas_dict = {}
if not rows.empty:
meas_dict = rows.to_dict('records')[0]
measurement_info = {
"sbp": meas_dict.get("sbp", ""),
"dbp": meas_dict.get("dbp", "")
}
for key, value in measurement_info.items():
hf.attrs[f"measurement_{key}"] = value
# 存入特徵資訊 (從 features.tsv 取得)
if features_df is not None:
feat_rows = features_df[
(features_df['pid'] == pid) &
(features_df['phase'] == phase) &
(features_df['measurement'] == measurement_for_compare)
]
if not feat_rows.empty:
feat_dict = feat_rows.to_dict('records')[0]
hf.attrs["feature_baseline_sbp"] = feat_dict.get("baseline_sbp", "")
hf.attrs["feature_baseline_dbp"] = feat_dict.get("baseline_dbp", "")
hf.attrs["feature_delta_sbp"] = feat_dict.get("delta_sbp", "")
hf.attrs["feature_delta_dbp"] = feat_dict.get("delta_dbp", "")
# print(f"EKG.shape: {ekg_windows.shape}")
# print(f"PPG.shape: {ppg_windows.shape}")
# print(f"Tonometry.shape: {tono_windows.shape}")
# for attr_key in hf.attrs:
# print(f" {attr_key}: {hf.attrs[attr_key]}")
# input("Press Enter to continue...")
except Exception as e:
print(f"存檔 {h5_filename} 發生錯誤: {e}，跳過此檔案")
continue
print(f"已存 {h5_filename}（來源: {filename}，共 {ekg_windows.shape[0]} 個視窗）")
print("全部轉檔完成！")