import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from models.model_attention_mil import MIL_Attention_fc_mtl
import pdb
import os
import pandas as pd
from utils.utils import *
from utils.core_utils import EarlyStopping, Accuracy_Logger
from utils.file_utils import save_pkl, load_pkl
from sklearn.metrics import roc_auc_score, roc_curve, auc
import h5py
from models.resnet_custom import resnet50_baseline
import math
from sklearn.preprocessing import label_binarize
def initiate_model(args, ckpt_path=None):
print('Init Model')
model_dict = {"dropout": args.drop_out, 'n_classes': args.n_classes}
if args.model_size is not None and args.model_type in ['clam', 'attention_mil', 'clam_simple']:
model_dict.update({"size_arg": args.model_size})
if args.model_type =='clam':
raise NotImplementedError
elif args.model_type =='clam_simple':
raise NotImplementedError
elif args.model_type == 'attention_mil':
model = MIL_Attention_fc_mtl(**model_dict)
else: # args.model_type == 'mil'
raise NotImplementedError
#model.relocate()
print_network(model)
if ckpt_path is not None:
ckpt = torch.load(ckpt_path)
ckpt_clean = {}
for key in ckpt.keys():
if 'instance_loss_fn' in key:
continue
ckpt_clean.update({key.replace('.module', ''):ckpt[key]})
model.load_state_dict(ckpt_clean, strict=True)
model.relocate()
model.eval()
return model
def eval(dataset, args, ckpt_path):
model = initiate_model(args, ckpt_path)
print('Init Loaders')
loader = get_simple_loader(dataset, collate_fn='MIL_mtl')
results_dict = summary(model, loader, args)
print('test_error_task1: ', results_dict['test_error_task1'])
print('auc_task1: ', results_dict['auc_task1'])
print('test_error_task2: ', results_dict['test_error_task2'])
print('auc_task2: ', results_dict['auc_task2'])
print('test_error_task3: ', results_dict['test_error_task3'])
print('auc_task3: ', results_dict['auc_task3'])
return model, results_dict
# patient_results, test_error, auc, aucs, df
def infer(dataset, args, ckpt_path, class_labels, site_labels):
model = initiate_model(args, ckpt_path)
df = infer_dataset(model, dataset, args, class_labels, site_labels)
return model, df
# Code taken from pytorch/examples for evaluating topk classification on on ImageNet
def accuracy(output, target, topk=(1,)):
"""Computes the accuracy over the k top predictions for the specified values of k"""
with torch.no_grad():
maxk = max(topk)
batch_size = target.size(0)
_, pred = output.topk(maxk, 1, True, True)
pred = pred.t()
correct = pred.eq(target.view(1, -1).expand_as(pred))
res = []
for k in topk:
correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
res.append(correct_k.mul_(1.0 / batch_size))
return res
def summary(model, loader, args):
device=torch.device("cuda" if torch.cuda.is_available() else "cpu")
logger_task1 = Accuracy_Logger(n_classes=args.n_classes[0])
logger_task2 = Accuracy_Logger(n_classes=args.n_classes[1])
logger_task3 = Accuracy_Logger(n_classes=args.n_classes[2])
model.eval()
test_error_task1 = 0.
test_loss_task1 = 0.
test_error_task2 = 0.
test_loss_task2 = 0.
test_error_task3 = 0.
test_loss_task3 = 0.
all_probs_task1 = np.zeros((len(loader), args.n_classes[0]))
all_labels_task1 = np.zeros(len(loader))
all_probs_task2 = np.zeros((len(loader), args.n_classes[1]))
all_labels_task2 = np.zeros(len(loader))
all_probs_task3 = np.zeros((len(loader), args.n_classes[2]))
all_labels_task3 = np.zeros(len(loader))
if not args.patient_level:
slide_ids = loader.dataset.slide_data['slide_id']
patient_results = {}
for batch_idx, (data, label_task1, label_task2, label_task3) in enumerate(loader):
data = data.to(device)
label_task1 = label_task1.to(device)
label_task2 = label_task2.to(device)
label_task3 = label_task3.to(device)
slide_id = slide_ids.iloc[batch_idx]
with torch.no_grad():
model_results_dict = model(data)
logits_task1, Y_prob_task1, Y_hat_task1 = model_results_dict['logits_task1'], model_results_dict['Y_prob_task1'], model_results_dict['Y_hat_task1']
logits_task2, Y_prob_task2, Y_hat_task2 = model_results_dict['logits_task2'], model_results_dict['Y_prob_task2'], model_results_dict['Y_hat_task2']
logits_task3, Y_prob_task3, Y_hat_task3 = model_results_dict['logits_task3'], model_results_dict['Y_prob_task3'], model_results_dict['Y_hat_task3']
del model_results_dict
logger_task1.log(Y_hat_task1, label_task1)
logger_task2.log(Y_hat_task2, label_task2)
logger_task3.log(Y_hat_task3, label_task3)
probs_task1 = Y_prob_task1.cpu().numpy()
all_probs_task1[batch_idx] = probs_task1
all_labels_task1[batch_idx] = label_task1.item()
probs_task2 = Y_prob_task2.cpu().numpy()
all_probs_task2[batch_idx] = probs_task2
all_labels_task2[batch_idx] = label_task2.item()
probs_task3 = Y_prob_task3.cpu().numpy()
all_probs_task3[batch_idx] = probs_task3
all_labels_task3[batch_idx] = label_task3.item()
patient_results.update({slide_id: {'slide_id': np.array(slide_id),
'prob_task1': probs_task1, 'label_task1': label_task1.item(),
'prob_task2': probs_task2, 'label_task2': label_task2.item(),
'prob_task3': probs_task3, 'label_task3': label_task3.item() }})
error_task1 = calculate_error(Y_hat_task1, label_task1)
test_error_task1 += error_task1
error_task2 = calculate_error(Y_hat_task2, label_task2)
test_error_task2 += error_task2
error_task3 = calculate_error(Y_hat_task3, label_task3)
test_error_task3 += error_task3
else:
case_ids = loader.dataset.slide_data['case_id']
patient_results = {}
for batch_idx, (data, label_task1, label_task2, label_task3) in enumerate(loader):
data = data.to(device)
label_task1 = label_task1.to(device)
label_task2 = label_task2.to(device)
label_task3 = label_task3.to(device)
case_id = case_ids.iloc[batch_idx]
with torch.no_grad():
model_results_dict = model(data)
logits_task1, Y_prob_task1, Y_hat_task1 = model_results_dict['logits_task1'], model_results_dict['Y_prob_task1'], model_results_dict['Y_hat_task1']
logits_task2, Y_prob_task2, Y_hat_task2 = model_results_dict['logits_task2'], model_results_dict['Y_prob_task2'], model_results_dict['Y_hat_task2']
logits_task3, Y_prob_task3, Y_hat_task3 = model_results_dict['logits_task3'], model_results_dict['Y_prob_task3'], model_results_dict['Y_hat_task3']
del model_results_dict
logger_task1.log(Y_hat_task1, label_task1)
logger_task2.log(Y_hat_task2, label_task2)
logger_task3.log(Y_hat_task3, label_task3)
probs_task1 = Y_prob_task1.cpu().numpy()
all_probs_task1[batch_idx] = probs_task1
all_labels_task1[batch_idx] = label_task1.item()
probs_task2 = Y_prob_task2.cpu().numpy()
all_probs_task2[batch_idx] = probs_task2
all_labels_task2[batch_idx] = label_task2.item()
probs_task3 = Y_prob_task3.cpu().numpy()
all_probs_task3[batch_idx] = probs_task3
all_labels_task3[batch_idx] = label_task3.item()
patient_results.update({case_id: {'case_id': np.array(case_id),
'prob_task1': probs_task1, 'label_task1': label_task1.item(),
'prob_task2': probs_task2, 'label_task2': label_task2.item(),
'prob_task3': probs_task3, 'label_task3': label_task3.item() }})
error_task1 = calculate_error(Y_hat_task1, label_task1)
test_error_task1 += error_task1
error_task2 = calculate_error(Y_hat_task2, label_task2)
test_error_task2 += error_task2
error_task3 = calculate_error(Y_hat_task3, label_task3)
test_error_task3 += error_task3
test_error_task1 /= len(loader)
test_error_task2 /= len(loader)
test_error_task3 /= len(loader)
all_preds_task1 = np.argmax(all_probs_task1, axis=1)
all_preds_task2 = np.argmax(all_probs_task2, axis=1)
all_preds_task3 = np.argmax(all_probs_task3, axis=1)
#if args.n_classes > 2:
# acc1, acc3 = accuracy(torch.from_numpy(all_cls_probs), torch.from_numpy(all_cls_labels), topk=(1, 3))
# print('top1 acc: {:.3f}, top3 acc: {:.3f}'.format(acc1.item(), acc3.item()))
# IF MORE THAN BINARY CLASSIFICATION
#if len(np.unique(all_labels_task1)) == 1:
# auc_task1 = -1
# aucs_task1 = []
# else:
# if args.n_classes[0] == 2:
# auc_task1 = roc_auc_score(all_labels_task1, all_probs_task1[:, 1])
# aucs_task1 = []
# else:
# aucs_task1 = []
# binary_labels = label_binarize(all_labels_task1, classes=[i for i in range(args.n_classes[0])])
# for class_idx in range(args.n_classes[0[]]):
# if class_idx in all_labels_task1:
# fpr, tpr, _ = roc_curve(binary_labels[:, class_idx], all_probs_task1[:, class_idx])
# aucs_task1.append(auc(fpr, tpr))
# else:
# aucs_task1.append(float('nan'))
# if args.micro_average:
# binary_labels = label_binarize(all_labels_task1, classes=[i for i in range(args.n_classes[0])])
# fpr, tpr, _ = roc_curve(binary_labels.ravel(), all_probs_task1.ravel())
# auc_task1 = auc(fpr, tpr)
# else:
# auc_task1 = np.nanmean(np.array(aucs_task1))
# ASSUME BINARY CLASSIFICATION
if len(np.unique(all_labels_task1)) == 1:
auc_task1 = -1
else:
auc_task1 = roc_auc_score(all_labels_task1, all_probs_task1[:, 1])
if len(np.unique(all_labels_task2)) == 1:
auc_task2 = -1
else:
auc_task2 = roc_auc_score(all_labels_task2, all_probs_task2[:, 1])
if len(np.unique(all_labels_task3)) == 1:
auc_task3 = -1
else:
auc_task3 = roc_auc_score(all_labels_task3, all_probs_task3[:, 1])
if not args.patient_level:
results_dict = {'slide_id': slide_ids,
'Y_task1': all_labels_task1, 'Y_hat_task1': all_preds_task1,
'Y_task2': all_labels_task2, 'Y_hat_task2': all_preds_task2,
'Y_task3': all_labels_task3, 'Y_hat_task3': all_preds_task3}
else:
results_dict = {'case_id': case_ids,
'Y_task1': all_labels_task1, 'Y_hat_task1': all_preds_task1,
'Y_task2': all_labels_task2, 'Y_hat_task2': all_preds_task2,
'Y_task3': all_labels_task3, 'Y_hat_task3': all_preds_task3}
results_dict.update({'p0_task1': all_probs_task1[:,0]})
results_dict.update({'p1_task1': all_probs_task1[:,1]})
results_dict.update({'p0_task2': all_probs_task2[:,0]})
results_dict.update({'p1_task2': all_probs_task2[:,1]})
results_dict.update({'p0_task3': all_probs_task3[:,0]})
results_dict.update({'p1_task3': all_probs_task3[:,1]})
df = pd.DataFrame(results_dict)
if args.patient_level:
df = df.drop_duplicates(subset=['case_id'])
inference_results = {'patient_results': patient_results,
'test_error_task1': test_error_task1, 'auc_task1': auc_task1,
'test_error_task2': test_error_task2, 'auc_task2': auc_task2,
'test_error_task3': test_error_task3, 'auc_task3': auc_task3,
'loggers': (logger_task1, logger_task2, logger_task3), 'df':df}
return inference_results
def infer_dataset(model, dataset, args, class_labels, site_labels, k=3):
model.eval()
all_probs_cls = np.zeros((len(dataset), k))
all_probs_site = np.zeros((len(dataset),2))
all_preds_cls = np.zeros((len(dataset), k))
all_preds_cls_str = np.full((len(dataset), k), ' ', dtype=object)
all_preds_site = np.full((len(dataset)), ' ', dtype=object)
slide_ids = dataset.slide_data
for batch_idx, data in enumerate(dataset):
data = data.to(device)
with torch.no_grad():
results_dict = model(data)
Y_prob, Y_hat = results_dict['Y_prob'], results_dict['Y_hat']
site_prob, site_hat = results_dict['site_prob'], results_dict['site_hat']
del results_dict
probs, ids = torch.topk(Y_prob, k)
probs = probs.cpu().numpy()
site_prob = site_prob.cpu().numpy()
ids = ids.cpu().numpy()
all_probs_cls[batch_idx] = probs
all_preds_cls[batch_idx] = ids
all_preds_cls_str[batch_idx] = np.array(class_labels)[ids]
all_probs_site[batch_idx] = site_prob
all_preds_site[batch_idx] = np.array(site_labels)[site_hat.item()]
del data
results_dict = {'slide_id': slide_ids}
for c in range(k):
results_dict.update({'Pred_{}'.format(c): all_preds_cls_str[:, c]})
results_dict.update({'p_{}'.format(c): all_probs_cls[:, c]})
results_dict.update({'Site_Pred': all_preds_site, 'Site_p': all_probs_site[:, 1]})
df = pd.DataFrame(results_dict)
return df
# def infer_dataset(model, dataset, args, class_labels, k=3):
# model.eval()
# all_probs = np.zeros((len(dataset), args.n_classes))
# all_preds = np.zeros(len(dataset))
# all_str_preds = np.full(len(dataset), ' ', dtype=object)
# slide_ids = dataset.slide_data
# for batch_idx, data in enumerate(dataset):
# data = data.to(device)
# with torch.no_grad():
# logits, Y_prob, Y_hat, _, results_dict = model(data)
# probs = Y_prob.cpu().numpy()
# all_probs[batch_idx] = probs
# all_preds[batch_idx] = Y_hat.item()
# all_str_preds[batch_idx] = class_labels[Y_hat.item()]
# del data
# results_dict = {'slide_id': slide_ids, 'Prediction': all_str_preds, 'Y_hat': all_preds}
# for c in range(args.n_classes):
# results_dict.update({'p_{}_{}'.format(c, class_labels[c]): all_probs[:,c]})
# df = pd.DataFrame(results_dict)
# return df
def compute_features(dataset, args, ckpt_path, save_dir, model=None, feature_dim=512):
if model is None:
model = initiate_model(args, ckpt_path)
names = dataset.get_list(np.arange(len(dataset))).values
file_path = os.path.join(save_dir, 'features.h5')
initialize_features_hdf5_file(file_path, len(dataset), feature_dim=feature_dim, names=names)
for i in range(len(dataset)):
print("Progress: {}/{}".format(i, len(dataset)))
save_features(dataset, i, model, args, file_path)
def save_features(dataset, idx, model, args, save_file_path):
name = dataset.get_list(idx)
print(name)
features, label_task1, label_task2, label_task3 = dataset[idx]
features = features.to(device)
with torch.no_grad():
results_dict = model(features, return_features=True)
Y_prob_task1, Y_hat_task1 = results_dict['Y_prob_task1'], results_dict['Y_hat_task1']
Y_prob_task2, Y_hat_task2 = results_dict['Y_prob_task2'], results_dict['Y_hat_task2']
Y_prob_task3, Y_hat_task3 = results_dict['Y_prob_task3'], results_dict['Y_hat_task3']
feat_task1 = results_dict['features'][0]
feat_task2 = results_dict['features'][1]
feat_task3 = results_dict['features'][2]
del results_dict
del features
Y_hat_task1 = Y_hat_task1.item()
Y_prob_task1 = Y_prob_task1.view(-1).cpu().numpy()
Y_hat_task2 = Y_hat_task2.item()
Y_prob_task2 = Y_prob_task2.view(-1).cpu().numpy()
Y_hat_task3 = Y_hat_task3.item()
Y_prob_task3 = Y_prob_task3.view(-1).cpu().numpy()
feat_task1 = feat_task1.view(1, -1).cpu().numpy()
feat_task2 = feat_task2.view(1, -1).cpu().numpy()
feat_task3 = feat_task3.view(1, -1).cpu().numpy()
with h5py.File(save_file_path, 'r+') as file:
print('label_task1', label_task1)
file['features_task1'][idx, :] = feat_task1
file['features_task2'][idx, :] = feat_task2
file['features_task3'][idx, :] = feat_task3
file['label_task1'][idx] = label_task1
file['Y_hat_task1'][idx] = Y_hat_task1
file['Y_prob_task1'][idx] = Y_prob_task1[1]
file['label_task2'][idx] = label_task2
file['Y_hat_task2'][idx] = Y_hat_task2
file['Y_prob_task2'][idx] = Y_prob_task2[1]
file['label_task3'][idx] = label_task3
file['Y_hat_task3'][idx] = Y_hat_task3
file['Y_prob_task3'][idx] = Y_prob_task3[1]
def initialize_features_hdf5_file(file_path, length, feature_dim=512, names = None):
file = h5py.File(file_path, "w")
dset = file.create_dataset('features_task1',
shape=(length, feature_dim), chunks=(1, feature_dim), dtype=np.float32)
dset = file.create_dataset('features_task2',
shape=(length, feature_dim), chunks=(1, feature_dim), dtype=np.float32)
dset = file.create_dataset('features_task3',
shape=(length, feature_dim), chunks=(1, feature_dim), dtype=np.float32)
# if names is not None:
# names = np.array(names, dtype='S')
# dset.attrs['names'] = names
if names is not None:
dt = h5py.string_dtype()
label_dset = file.create_dataset('names', shape=(length, ), chunks=(1, ), dtype=dt)
file['names'][:] = names
label_dset = file.create_dataset('label_task1', shape=(length, ), chunks=(1, ), dtype=np.int32)
pred_dset = file.create_dataset( 'Y_hat_task1', shape=(length, ), chunks=(1, ), dtype=np.int32)
prob_dset = file.create_dataset( 'Y_prob_task1', shape=(length, ), chunks=(1, ), dtype=np.float32)
label_dset = file.create_dataset('label_task2', shape=(length, ), chunks=(1, ), dtype=np.int32)
pred_dset = file.create_dataset( 'Y_hat_task2', shape=(length, ), chunks=(1, ), dtype=np.int32)
prob_dset = file.create_dataset( 'Y_prob_task2', shape=(length, ), chunks=(1, ), dtype=np.float32)
label_dset = file.create_dataset('label_task3', shape=(length, ), chunks=(1, ), dtype=np.int32)
pred_dset = file.create_dataset( 'Y_hat_task3', shape=(length, ), chunks=(1, ), dtype=np.int32)
prob_dset = file.create_dataset( 'Y_prob_task3', shape=(length, ), chunks=(1, ), dtype=np.float32)
file.close()
return file_path
Datasets
Models
