Add synthetic waveform metadata

data/generate_synthetic_waveform_metadata.py

@@ -0,0 +1,137 @@
+import csv
+from datetime import datetime, timedelta
+
+N_SUBJECTS = 20
+BASE_YEAR_REAL = 2045
+BASE_YEAR_MIMIC = 2065
+
+CHANNEL_PAIRS = [
+    ("PLETH", "NU", 250, "adu/NU"),
+    ("RESP", "Ohm", 210, "adu/Ohm"),
+]
+
+ECG_LEADS = ["II", "III", "V", "aVR"]
+
+def dt(y, m, d, h=0, mi=0, s=0):
+    return datetime(y, m, d, h, mi, s)
+
+def fmt(t):
+    return t.strftime("%Y-%m-%d %H:%M:%S.000000 UTC")
+
+def choose_channels(i, segment_idx):
+    base = CHANNEL_PAIRS[(i + segment_idx) % len(CHANNEL_PAIRS)]
+    ecg = ECG_LEADS[(i + segment_idx) % len(ECG_LEADS)]
+    return [
+        (base[0], base[1], base[2], base[3]),
+        (ecg, "mV", 125, "adu/mV"),
+    ]
+
+waveform_files = []
+waveform_channels = []
+
+for i in range(1, N_SUBJECTS + 1):
+    subject_id = 20000000 + i
+    person_id = 9000000000 + i
+
+    # decide structure
+    n_recordings = 2 if i in (4, 10) else 1
+    multi_segment = i in (3, 5)
+
+    for r in range(n_recordings):
+        rec_idx = i * 1000 + r
+        hadm_id = 84000000 + rec_idx
+        visit_occ = 9500000000 + rec_idx
+        visit_detail = 9600000000 + rec_idx
+        group_id = 9001000 + rec_idx
+
+        folder_prefix = 80 + (i % 5)
+        waveform_folder = f"{folder_prefix}/{group_id}"
+        location = ["icu", "ccu"][i % 2]
+
+        t0_real = dt(BASE_YEAR_REAL + i % 5, i % 12 + 1, i % 28 + 1, i % 24)
+        t0_mimic = dt(BASE_YEAR_MIMIC + i % 5, i % 12 + 1, i % 28 + 1, i % 24)
+
+        session_duration = timedelta(hours=8)
+        session_end_real = t0_real + session_duration
+        session_end_mimic = t0_mimic + session_duration
+
+        # --- master header ---
+        waveform_files.append([
+            subject_id, person_id, hadm_id, visit_occ, visit_detail,
+            location, waveform_folder, group_id, group_id,
+            fmt(t0_mimic), fmt(session_end_mimic),
+            fmt(t0_real), fmt(session_end_real),
+            fmt(t0_real), fmt(session_end_real - timedelta(minutes=1)),
+            f"{waveform_folder}/{group_id}.hea",
+            f"{person_id}/Waveforms/{group_id}/{group_id}.hea"
+        ])
+
+        # segments
+        n_segments = 2 if multi_segment else 1
+
+        seg_start = t0_real + timedelta(minutes=1)
+
+        for s in range(1, n_segments + 1):
+            seg_len_sec = 60 + (s * 30)  # 60s, 90s
+            seg_end = seg_start + timedelta(seconds=seg_len_sec)
+
+            seg_name = f"{group_id}_{s:04d}"
+
+            for ext in ["hea", "dat"]:
+                waveform_files.append([
+                    subject_id, person_id, hadm_id, visit_occ, visit_detail,
+                    location, waveform_folder, group_id, group_id,
+                    fmt(t0_mimic), fmt(session_end_mimic),
+                    fmt(seg_start), fmt(seg_end),
+                    fmt(seg_start), fmt(seg_end),
+                    f"{waveform_folder}/{seg_name}.{ext}",
+                    f"{person_id}/Waveforms/{group_id}/{seg_name}.{ext}"
+                ])
+
+            # channels (4 rows per segment)
+            segment_length = int(seg_len_sec * 125)
+
+            channels = choose_channels(i, s)
+
+            for ext in ["dat", "hea"]:
+                for channel_index, (ch_name, unit, gain, gain_unit) in enumerate(channels):
+                    waveform_channels.append([
+                        person_id, visit_occ, visit_detail, group_id,
+                        fmt(t0_real), fmt(session_end_real),
+                        fmt(seg_start), fmt(seg_end),
+                        f"{waveform_folder}/{seg_name}.{ext}",
+                        f"{person_id}/Waveforms/{group_id}/{seg_name}.{ext}",
+                        channel_index,
+                        ch_name,
+                        unit,
+                        125,
+                        "Hz",
+                        gain,
+                        gain_unit,
+                        segment_length
+                    ])
+
+            seg_start = seg_end  # next segment starts immediately
+
+# --- write files ---
+with open("waveform_files.csv", "w", newline="") as f:
+    writer = csv.writer(f)
+    writer.writerow([
+        "subject_id","person_id","hadm_id","visit_occurrence_id","visit_detail_id",
+        "location","waveform_folders","record_id","group_id",
+        "mimic_start","mimic_end","session_start","session_end",
+        "file_start","file_end","src_file","trg_file"
+    ])
+    writer.writerows(waveform_files)
+
+with open("waveform_channels.csv", "w", newline="") as f:
+    writer = csv.writer(f)
+    writer.writerow([
+        "person_id","visit_occurrence_id","visit_detail_id","group_id",
+        "session_start","session_end","file_start","file_end",
+        "src_file","trg_file","channel_index","channel_name","sample_units",
+        "sample_rate","sample_rate_units","gain","gain_units","segment_length"
+    ])
+    writer.writerows(waveform_channels)
+
+print("Generated waveform_files.csv and waveform_channels.csv")