make_acc_splits.py

import argparse
import os
import re
import random

TCGA_ID_RE = re.compile(r"(TCGA-[A-Z0-9]{2}-[A-Z0-9]+)")
DX_TOKEN_RE = re.compile(r"(?:^|-)DX\d*(?=[\-.]|$)")
TS_TOKEN_RE = re.compile(r"(?:^|-)TS[A-Za-z0-9]*(?=[\-.]|$)")

def extract_case_id(name):
    m = TCGA_ID_RE.search(name)
    return m.group(1) if m else None

def is_dx_item(name):
    if TS_TOKEN_RE.search(name):
        return False
    return DX_TOKEN_RE.search(name) is not None

def collect_features(path):
    case_to_items = {}
    for n in os.listdir(path):
        if n.startswith('.'):
            continue
        if not n.endswith('.pt'):
            continue
        if not is_dx_item(n):
            continue
        cid = extract_case_id(n)
        if not cid:
            continue
        case_to_items.setdefault(cid, []).append(n)
    for k in list(case_to_items.keys()):
        case_to_items[k].sort()
    return case_to_items

def collect_label_cases(path):
    cases = set()
    for n in os.listdir(path):
        if n.startswith('.'):
            continue
        p = os.path.join(path, n)
        if not os.path.isdir(p):
            continue
        cid = extract_case_id(n)
        if cid:
            cases.add(cid)
    return cases

def write_csv(train_items, val_items, test_items, out_path):
    max_len = max(len(train_items), len(val_items), len(test_items))
    with open(out_path, 'w', encoding='utf-8') as f:
        f.write(',train,val,test\n')
        for i in range(max_len):
            t = train_items[i] if i < len(train_items) else ''
            v = val_items[i] if i < len(val_items) else ''
            te = test_items[i] if i < len(test_items) else ''
            t = os.path.splitext(t)[0]
            v = os.path.splitext(v)[0]
            te = os.path.splitext(te)[0]
            f.write(f"{i},{t},{v},{te}\n")

def main():
    ap = argparse.ArgumentParser()
    ap.add_argument('--feature_source', type=str, default='/mnt/datadisk0/TCGA_pt/MESO_UNI')
    ap.add_argument('--label_source', type=str, default='/mnt/datadisk0/datasets/TCGA-MESO')
    ap.add_argument('--output_csv', type=str, default='/mnt/datadisk0/BiGen/ocr/dataset_csv/splits_MESO.csv')
    ap.add_argument('--seed', type=int, default=42)
    args = ap.parse_args()

    features_map = collect_features(args.feature_source)
    label_cases = collect_label_cases(args.label_source)
    common_cases = sorted(set(features_map.keys()) & label_cases)

    random.seed(args.seed)
    random.shuffle(common_cases)

    total = len(common_cases)
    n_train = int(total * 0.8)
    n_val = int(total * 0.1)
    n_test = total - n_train - n_val

    train_cases = common_cases[:n_train]
    val_cases = common_cases[n_train:n_train + n_val]
    test_cases = common_cases[n_train + n_val:]

    train_items = []
    val_items = []
    test_items = []

    for c in train_cases:
        train_items.extend(features_map.get(c, []))
    for c in val_cases:
        val_items.extend(features_map.get(c, []))
    for c in test_cases:
        test_items.extend(features_map.get(c, []))

    write_csv(train_items, val_items, test_items, args.output_csv)

    print('Feature source:', args.feature_source)
    print('Label source:', args.label_source)
    print('Common cases:', total)
    print('Train cases:', len(train_cases))
    print('Val cases:', len(val_cases))
    print('Test cases:', len(test_cases))
    print('Train items:', len(train_items))
    print('Val items:', len(val_items))
    print('Test items:', len(test_items))
    print('Output CSV:', args.output_csv)

if __name__ == '__main__':
    main()