Code fragments of isicdataset.py

• from fhirpy import SyncFHIRClient
• import requests
• from io import BytesIO
• import numpy as np
• import pandas as pd
• from PIL import Image
• from torch.utils.data import Dataset
• import torchvision.transforms as transforms
• class ISICDataset(Dataset):
• classes = {'NV': 0, 'MEL': 1, 'BKL': 2, 'DF': 3, 'SCC': 4, 'BCC': 5, 'VASC': 6, 'AK': 7}
• """ISIC dataset."""
• def __init__(self, fhir_server, fhir_port, split='train', input_size=256):
• """ Args: fhir_server (string): Address of FHIR Server. fhir_port (string): Port of FHIR Server. """
• self.fhir_server = fhir_server
• self.fhir_port = fhir_port
• self.split = split
• self.input_size = input_size
• # Create an instance
• client = SyncFHIRClient('http://{}:{}/fhir'.format(fhir_server, fhir_port))
• # Search for patients
• # Return lazy search set
• patients = client.resources('Patient')
• patients_data = []
• for patient in patients:
• patient_birthDate = None
• try:
• patient_birthDate = patient.birthDate
• except: pass
• # patinet_id, gender, birthDate
• patients_data.append([patient.id, patient.gender, patient_birthDate])
• patients_df = pd.DataFrame(patients_data, columns=["patient_id", "gender", "birthDate"])
• # Search for media
• media_list = client.resources('Media').include('Patient', 'subject')
• media_data = []
• for media in media_list:
• media_bodySite = None
• media_reasonCode = None
• media_note = None
• try:
• media_bodySite = media.bodySite.text
• except: pass
• try:
• media_reasonCode = media.reasonCode[0].text
• except: pass
• try:
• media_note = media.note[0].text
• except: pass
• media_data.append([media.subject.id, media.id, media_bodySite, media_reasonCode, media_note, media.content.url])
• media_df = pd.DataFrame(media_data, columns=["patient_id", "media_id", "bodySite", "reasonCode", "note", "image_url"])
• self.data_df = pd.merge(patients_df, media_df, on='patient_id', how='outer')
• self.data_df = self.data_df[self.data_df['note'].notna()].reset_index()
• self.trans = transforms.Compose([transforms.RandomHorizontalFlip(), transforms.RandomVerticalFlip(), transforms.ColorJitter(brightness=32. / 255., saturation=0.5), transforms.Resize(self.input_size), transforms.ToTensor()])
• self.trans = transforms.Compose([transforms.RandomHorizontalFlip(), transforms.RandomVerticalFlip(), transforms.ColorJitter(brightness=32. / 255., saturation=0.5), transforms.Resize(self.input_size), transforms.ToTensor()])
• def __len__(self):
• if self.split == "train":
• return int(len(self.data_df) * 0.8)
• elif self.split == "val":
• return int(len(self.data_df) * 0.2)
• def __getitem__(self, idx):
• val_start_id = int(len(self.data_df) * 0.8)
• if self.split == "train":
• idx = idx
• else:
• idx = idx + val_start_id
• img_url = self.data_df.loc[idx, 'image_url']
• note = self.data_df.loc[idx, 'note']
• y = self.classes[note]
• img_res = requests.get(img_url)
• if img_res.status_code == 200:
• x = Image.open(BytesIO(img_res.content))
• x = self.center_crop(x)
• x = self.trans(x)
• # Transform y
• y = np.int64(y)
• return {"image": x, "label": y}
• else:
• raise RuntimeError("Image url {} is not reachable!".format(img_url))
• def center_crop(self, pil_img):
• img_width, img_height = pil_img.size
• if img_width > img_height:
• crop_size = img_height
• else:
• crop_size = img_width
• return pil_img.crop(((img_width - crop_size) // 2, (img_height - crop_size) // 2, (img_width + crop_size) // 2, (img_height + crop_size) // 2))
• return pil_img.crop(((img_width - crop_size) // 2, (img_height - crop_size) // 2, (img_width + crop_size) // 2, (img_height + crop_size) // 2))

Graph