import pdb
import os
import pandas as pd
import numpy as np
from datasets.dataset_generic import Generic_WSI_Classification_Dataset, Generic_MIL_Dataset, save_splits
import argparse
parser = argparse.ArgumentParser(description='Creating splits for whole slide classification')
parser.add_argument('--label_frac', type=float, default= 1,
help='fraction of labels (default: [0.25, 0.5, 0.75, 1.0])')
parser.add_argument('--seed', type=int, default=1,
help='random seed (default: 1)')
parser.add_argument('--k', type=int, default=10,
help='number of splits (default: 10)')
parser.add_argument('--task', type=str, choices=['cardiac-grade','cardiac-mtl'],
help='select one of the supported tasks for which to perform the split')
parser.add_argument('--hold_out_test', action='store_true', default=False,
help='hold-out the test set for each split')
parser.add_argument('--split_code', type=str, default=None)
args = parser.parse_args()
# The splits are made to account for all possible diagnosis across all splits:
# grade - cell rejection only low grade
# - low grade cellular, high-grade antibody case ...
# - ....
# mtl - low grade cellular only
# - high grade cellular only
# - low grade cellular + quilty
# - .....
#---------------------------------------------------
if args.task == 'cardiac-grade':
dataset = Generic_WSI_Classification_Dataset(csv_path = 'dataset_csv/CardiacDummy_GradeSplit.csv',
shuffle = False,
seed = args.seed,
print_info = True,
label_dict = {'cell_only_low' : 0,
'cell_only_high' : 1,
'cell_low_quilty' : 2,
'cell_high_quilty' : 3,
'amr_only_low' : 4,
'amr_only_high' : 5,
'amr_low_quilty' : 6,
'amr_high_quilty' : 7,
'cell_amr_low' : 8,
'cell_amr_high' : 9,
'cell_amr_quilty_low' : 10,
'cell_amr_quilty_high' : 11},
patient_strat= True,
ignore=[])
p_val = 0.1 # use 10% of data in validation
p_test = 0.2 # to use hold-out test set set p_test = 0
elif args.task == 'cardiac-mtl':
dataset = Generic_WSI_Classification_Dataset(csv_path = 'dataset_csv/CardiacDummy_MTLSplit.csv',
shuffle = False,
seed = args.seed,
print_info = True,
label_dict = {'healthy' :0,
'quilty' :1,
'cell_only_low' :2,
'cell_only_high' :3,
'cell_low_quilty' :4,
'cell_high_quilty' :5,
'amr_only_low' :6,
'amr_only_high' :7,
'amr_low_quilty' :8,
'amr_high_quilty' :9,
'cell_amr_low' :10,
'cell_amr_high' :11,
'cell_amr_quilty_low' :12,
'cell_amr_quilty_high' :13},
patient_strat= True,
ignore=[])
p_val = 0.1 # use 10% of data in validation
p_test = 0.2 # use 20% data for test set
else:
raise NotImplementedError
# splits
num_slides_cls = np.array([len(cls_ids) for cls_ids in dataset.patient_cls_ids])
val_num = np.floor(num_slides_cls * p_val).astype(int) # use 10% data in validation
test_num = np.floor(num_slides_cls * p_test).astype(int) # use 20% for test set
print("---------------------------------")
print(f"validation set size = {val_num} ")
print(f"test set size = {test_num}")
print("---------------------------------")
if __name__ == '__main__':
if args.label_frac > 0:
label_fracs = [args.label_frac]
else:
label_fracs = [0.25, 0.5, 0.75, 1.0]
if args.hold_out_test:
custom_test_ids = dataset.sample_held_out(test_num=test_num)
else:
custom_test_ids = None
for lf in label_fracs:
if args.split_code is not None:
split_dir = 'splits/'+ str(args.split_code) + '_{}'.format(int(lf * 100))
else:
split_dir = 'splits/'+ str(args.task) + '_{}'.format(int(lf * 100))
os.makedirs(split_dir, exist_ok=True)
#pdb.set_trace()
dataset.create_splits(k = args.k, val_num = val_num, test_num = test_num, label_frac=lf, custom_test_ids=custom_test_ids)
for i in range(args.k):
dataset.set_splits()
descriptor_df = dataset.test_split_gen(return_descriptor=True)
splits = dataset.return_splits(from_id=True)
save_splits(splits, ['train', 'val','test'], os.path.join(split_dir, 'splits_{}.csv'.format(i)))
save_splits(splits, ['train', 'val','test'], os.path.join(split_dir, 'splits_{}_bool.csv'.format(i)), boolean_style=True)
descriptor_df.to_csv(os.path.join(split_dir, 'splits_{}_descriptor.csv'.format(i)))
Datasets
Models
