-
Notifications
You must be signed in to change notification settings - Fork 5
/
dsi.py
63 lines (48 loc) · 2.07 KB
/
dsi.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
import os
import tensorflow as tf
import numpy as np
from utility import _create_patient_data_df, _get_file_names, _parse_dicom_data
class OSICFibrosisDSI:
def __init__(self, csv_path, ct_path):
self.csv_path = csv_path
self.ct_path = ct_path
def _load_data(self):
self._test_patient_data_df = _create_patient_data_df(self.csv_path)
all_files = _get_file_names(self.ct_path)
dataset = tf.data.Dataset.from_tensor_slices(all_files)
num_data = len(all_files)
dataset = dataset.map(_parse_dicom_data)
dataset = dataset.map(lambda x: self._add_patient_data_map(x))
return dataset, num_data
def get_test_dataset(self):
batch_size = 1
dataset, num_data = self._load_data()
dataset = dataset.batch(batch_size)
# use 1 for test batch size
return dataset, num_data, batch_size
def _get_patient_data(self, patient_id):
""" Gets wrapped in tf.numpy_func, gets patient data
"""
patient_id = patient_id.decode("utf-8")
patient_df = self._test_patient_data_df
data = patient_df[
patient_df.Patient == patient_id
].iloc[0][1:].values.tolist()
# Ensure all elements are arrays so they can be concatenated
for i in range(len(data)):
if isinstance(data[i], np.ndarray):
pass
else:
data[i] = np.array([data[i]])
data = np.concatenate(data).astype(np.float32)
return data
def _add_patient_data_map(self, data_tsr):
""" Adds patient metadata to tsr containing just patient_id and image
"""
data_vec = tf.numpy_function(
self._get_patient_data, [data_tsr['Patient']], Tout=tf.float32)
data_vec = tf.split(data_vec, [4,1,1])
data_tsr['metadata'] = tf.ensure_shape(data_vec[0], [4])
data_tsr['cur_fvc'] = tf.ensure_shape(data_vec[1], [1])
data_tsr['cur_week'] = tf.ensure_shape(data_vec[2], [1])
return data_tsr