from __future__ import print_function
import argparse
import pdb
import os
import math
import sys
# internal imports
from utils.file_utils import save_pkl, load_pkl
from utils.utils import *
from utils.core_utils import train, eval_model
from datasets.dataset_generic import Generic_WSI_Classification_Dataset, Generic_MIL_Dataset
from datasets.dataset_survival import Generic_WSI_Survival_Dataset, Generic_MIL_Survival_Dataset
# pytorch imports
import torch
from torch.utils.data import DataLoader, sampler
import torch.nn as nn
import torch.nn.functional as F
import pandas as pd
import numpy as np
from timeit import default_timer as timer
def main(args):
# create results directory if necessary
if not os.path.isdir(args.results_dir):
os.mkdir(args.results_dir)
if args.k_start == -1:
start = 0
else:
start = args.k_start
if args.k_end == -1:
end = args.k
else:
end = args.k_end
val_cindex = []
folds = np.arange(start, end)
for i in folds:
start = timer()
seed_torch(args.seed)
train_dataset, val_dataset = dataset.return_splits(from_id=False, csv_path='{}/splits_{}.csv'.format(args.split_dir, i))
print('training: {}, validation: {}'.format(len(train_dataset), len(val_dataset)))
datasets = (train_dataset, val_dataset)
if 'omic' in args.mode:
args.omic_input_dim = train_dataset.genomic_features.shape[1]
print("Genomic Dimension", args.omic_input_dim)
val_latest, cindex_latest = eval_model(datasets, i, args)
val_cindex.append(cindex_latest)
#write results to pkl
save_pkl(os.path.join(args.results_dir, 'split_val_{}_results.pkl'.format(i)), val_latest)
end = timer()
print('Fold %d Time: %f seconds' % (i, end - start))
if len(folds) != args.k: save_name = 'summary_partial_{}_{}.csv'.format(start, end)
else: save_name = 'summary.csv'
results_df = pd.DataFrame({'folds': folds, 'val_cindex': val_cindex})
results_df.to_csv(os.path.join(args.results_dir, 'summary.csv'))
# Training settings
parser = argparse.ArgumentParser(description='Configurations for MMF Training')
parser.add_argument('--data_root_dir', type=str, default='/media/ssd1/pan-cancer', help='data directory')
parser.add_argument('--which_splits', type=str, default='5foldcv', help='Path to splits directory.')
parser.add_argument('--split_dir', type=str, help='Set of splits to use for each cancer type.')
parser.add_argument('--mode', type=str, default='omic')
parser.add_argument('--model_type', type=str, default='clam', help='type of model (attention_mil | max_net | mm_attention_mil)')
parser.add_argument('--max_epochs', type=int, default=20, help='maximum number of epochs to train (default: 20)')
parser.add_argument('--lr', type=float, default=2e-4, help='learning rate (default: 0.0001)')
parser.add_argument('--label_frac', type=float, default=1.0, help='fraction of training labels (default: 1.0)')
parser.add_argument('--bag_weight', type=float, default=0.7, help='clam: weight coefficient for bag-level loss (default: 0.7)')
parser.add_argument('--reg', type=float, default=1e-5, help='weight decay (default: 1e-5)')
parser.add_argument('--seed', type=int, default=1, help='random seed for reproducible experiment (default: 1)')
parser.add_argument('--k', type=int, default=5, help='number of folds (default: 10)')
parser.add_argument('--k_start', type=int, default=-1, help='start fold (default: -1, last fold)')
parser.add_argument('--k_end', type=int, default=-1, help='end fold (default: -1, first fold)')
parser.add_argument('--results_dir', default='./results', help='results directory (default: ./results)')
parser.add_argument('--log_data', action='store_true', default=True, help='log data using tensorboard')
parser.add_argument('--testing', action='store_true', default=False, help='debugging tool')
parser.add_argument('--early_stopping', action='store_true', default=False, help='enable early stopping')
parser.add_argument('--opt', type=str, choices = ['adam', 'sgd'], default='adam')
parser.add_argument('--drop_out', action='store_true', default=True, help='enabel dropout (p=0.25)')
parser.add_argument('--inst_loss', type=str, choices=['svm', 'ce', None], default=None, help='instance-level clustering loss function (default: None)')
parser.add_argument('--bag_loss', type=str, choices=['svm', 'ce', 'ce_surv', 'nll_surv', 'cox_surv'], default='nll_surv', help='slide-level classification loss function (default: ce)')
parser.add_argument('--alpha_surv', type=float, default=0.0, help='How much to weigh uncensored patients')
parser.add_argument('--reg_type', type=str, choices=['None', 'omic', 'pathomic'], default='None', help='Reg Type (default: None)')
parser.add_argument('--lambda_reg', type=float, default=1e-4, help='Regularization Strength')
parser.add_argument('--weighted_sample', action='store_true', default=True, help='enable weighted sampling')
parser.add_argument('--model_size_wsi', type=str, default='small', help='Size of AMIL model.')
parser.add_argument('--model_size_omic', type=str, default='small', help='Size of SNN Model.')
parser.add_argument('--gc', type=int, default=1, help='gradient accumulation step')
parser.add_argument('--batch_size', type=int, default=1, help='Batch Size')
parser.add_argument('--gate_path', action='store_true', default=False, help='Enable feature gating in MMF layer.')
parser.add_argument('--gate_omic', action='store_true', default=False, help='Enable feature gating in MMF layer.')
parser.add_argument('--fusion', type=str, default='tensor', help='Which fusion mechanism to use.')
parser.add_argument('--overwrite', action='store_true', default=False, help='Current experiment results already exists. Redo?')
parser.add_argument('--apply_mad', action='store_true', default=True, help='Use genes with median absolute deviation.')
parser.add_argument('--task', type=str, default='survival', help='Which task.')
args = parser.parse_args()
device=torch.device("cuda" if torch.cuda.is_available() else "cpu")
### Creates Custom Experiment Code
exp_code = '_'.join(args.split_dir.split('_')[:2])
dataset_path = 'dataset_csv'
param_code = ''
if args.model_type == 'attention_mil':
param_code += 'WSI'
elif args.model_type == 'max_net':
param_code += 'SNN'
elif args.model_type == 'mm_attention_mil' and args.fusion == 'tensor':
param_code += 'MMF'
else:
raise NotImplementedError
if 'small' in args.model_size_omic:
param_code += 'sm'
param_code += '_%s' % args.bag_loss
if 'mm_' in args.model_type:
param_code += '_g'
if args.gate_path:
param_code += '1'
else:
param_code += '0'
if args.gate_omic:
param_code += '1'
else:
param_code += '0'
param_code += '_a%s' % str(args.alpha_surv)
if args.lr != 2e-4:
param_code += '_lr%s' % format(args.lr, '.0e')
if args.reg_type != 'None':
param_code += '_reg%s' % format(args.lambda_reg, '.0e')
param_code += '_%s' % args.which_splits.split("_")[0]
if args.gc != 1:
param_code += '_gc%s' % str(args.gc)
if args.apply_mad:
param_code += '_mad'
#dataset_path += '_mad'
args.exp_code = exp_code + "_" + param_code
### task
if args.task == 'survival':
args.task = '_'.join(args.split_dir.split('_')[:2]) + '_survival'
print("Experiment Name:", exp_code)
def seed_torch(seed=7):
import random
random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
if device.type == 'cuda':
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed) # if you are using multi-GPU.
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True
seed_torch(args.seed)
encoding_size = 1024
settings = {'num_splits': args.k,
'k_start': args.k_start,
'k_end': args.k_end,
'task': args.task,
'max_epochs': args.max_epochs,
'results_dir': args.results_dir,
'lr': args.lr,
'experiment': args.exp_code,
'reg': args.reg,
'label_frac': args.label_frac,
'inst_loss': args.inst_loss,
'bag_loss': args.bag_loss,
'bag_weight': args.bag_weight,
'seed': args.seed,
'model_type': args.model_type,
'model_size_wsi': args.model_size_wsi,
'model_size_omic': args.model_size_omic,
"use_drop_out": args.drop_out,
'weighted_sample': args.weighted_sample,
'gc': args.gc,
'opt': args.opt}
print('\nLoad Dataset')
if args.task == 'tcga_blca_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_bladder_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_brca_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_breast_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_coadread_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_coadread_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_gbmlgg_survival':
args.n_classes = 4
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/tcga_gbmlgg_all.csv' % dataset_path,
mode = args.mode,
data_dir={'ASTR': os.path.join(args.data_root_dir,'tcga_lgg_20x_features'),
'AASTR': os.path.join(args.data_root_dir,'tcga_lgg_20x_features'),
'ODG': os.path.join(args.data_root_dir,'tcga_lgg_20x_features'),
'OAST': os.path.join(args.data_root_dir,'tcga_lgg_20x_features'),
'AOAST': os.path.join(args.data_root_dir,'tcga_lgg_20x_features'),
'GBM': os.path.join(args.data_root_dir,'tcga_gbm_20x_features'),},
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_hnsc_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_hnsc_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_kirc_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_kidney_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_kirp_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_kidney_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_lihc_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_liver_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_luad_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_lung_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_lusc_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_lung_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_paad_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_pancreas_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_skcm_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_skin_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_stad_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_stomach_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
elif args.task == 'tcga_ucec_survival':
args.n_classes = 4
proj = '_'.join(args.task.split('_')[:2])
dataset = Generic_MIL_Survival_Dataset(csv_path = './%s/%s_all.csv' % (dataset_path, proj),
mode = args.mode,
data_dir= os.path.join(args.data_root_dir, 'tcga_endometrial_20x_features'),
shuffle = False,
seed = args.seed,
print_info = True,
patient_strat= False,
n_bins=4,
label_col = 'survival_months',
ignore=[])
else:
raise NotImplementedError
if isinstance(dataset, Generic_MIL_Survival_Dataset):
args.task_type ='survival'
else:
raise NotImplementedError
if not os.path.isdir(args.results_dir):
os.mkdir(args.results_dir)
### GET RID OF WHICH_SPLITS IF U WANT TO MAKE THE RESULTS FOLDER LESS CLUTTERRED
args.results_dir = os.path.join(args.results_dir, args.which_splits, param_code, str(args.exp_code) + '_s{}'.format(args.seed))
if not os.path.isdir(args.results_dir):
os.makedirs(args.results_dir)
if ('summary.csv' in os.listdir(args.results_dir)) and (not args.overwrite):
print("Exp Code <%s> already exists! Exiting script." % args.exp_code)
sys.exit()
if args.split_dir is None:
args.split_dir = os.path.join('./splits', args.task+'_{}'.format(int(args.label_frac*100)))
else:
args.split_dir = os.path.join('./splits', args.which_splits, args.split_dir)
print("split_dir", args.split_dir)
assert os.path.isdir(args.split_dir)
settings.update({'split_dir': args.split_dir})
with open(args.results_dir + '/experiment_{}.txt'.format(args.exp_code), 'w') as f:
print(settings, file=f)
f.close()
print("################# Settings ###################")
for key, val in settings.items():
print("{}: {}".format(key, val))
if __name__ == "__main__":
start = timer()
results = main(args)
end = timer()
print("finished!")
print("end script")
print('Script Time: %f seconds' % (end - start))
Datasets
Models
