Chris Oswald commited on
Commit
9997395
·
1 Parent(s): 18d6eb1

flattened radiological gradings

Browse files
Files changed (1) hide show
  1. SPIDER.py +29 -106
SPIDER.py CHANGED
@@ -37,6 +37,7 @@ def import_csv_data(filepath: str) -> List[Dict[str, str]]:
37
 
38
  # Define constants
39
  N_PATIENTS = 257
 
40
  MAX_IVD = 9
41
 
42
  # TODO: Add BibTeX citation
@@ -181,96 +182,15 @@ class SPIDER(datasets.GeneratorBasedBuilder):
181
  "WindowWidth": datasets.Value(dtype="string"),
182
  },
183
  "rad_gradings": {
184
- "IVD1": {
185
- "Modic": datasets.Value(dtype="string"),
186
- "UP endplate": datasets.Value(dtype="string"),
187
- "LOW endplate": datasets.Value(dtype="string"),
188
- "Spondylolisthesis": datasets.Value(dtype="string"),
189
- "Disc herniation": datasets.Value(dtype="string"),
190
- "Disc narrowing": datasets.Value(dtype="string"),
191
- "Disc bulging": datasets.Value(dtype="string"),
192
- "Pfirrman grade": datasets.Value(dtype="string"),
193
- },
194
- "IVD2": {
195
- "Modic": datasets.Value(dtype="string"),
196
- "UP endplate": datasets.Value(dtype="string"),
197
- "LOW endplate": datasets.Value(dtype="string"),
198
- "Spondylolisthesis": datasets.Value(dtype="string"),
199
- "Disc herniation": datasets.Value(dtype="string"),
200
- "Disc narrowing": datasets.Value(dtype="string"),
201
- "Disc bulging": datasets.Value(dtype="string"),
202
- "Pfirrman grade": datasets.Value(dtype="string"),
203
- },
204
- "IVD3": {
205
- "Modic": datasets.Value(dtype="string"),
206
- "UP endplate": datasets.Value(dtype="string"),
207
- "LOW endplate": datasets.Value(dtype="string"),
208
- "Spondylolisthesis": datasets.Value(dtype="string"),
209
- "Disc herniation": datasets.Value(dtype="string"),
210
- "Disc narrowing": datasets.Value(dtype="string"),
211
- "Disc bulging": datasets.Value(dtype="string"),
212
- "Pfirrman grade": datasets.Value(dtype="string"),
213
- },
214
- "IVD4": {
215
- "Modic": datasets.Value(dtype="string"),
216
- "UP endplate": datasets.Value(dtype="string"),
217
- "LOW endplate": datasets.Value(dtype="string"),
218
- "Spondylolisthesis": datasets.Value(dtype="string"),
219
- "Disc herniation": datasets.Value(dtype="string"),
220
- "Disc narrowing": datasets.Value(dtype="string"),
221
- "Disc bulging": datasets.Value(dtype="string"),
222
- "Pfirrman grade": datasets.Value(dtype="string"),
223
- },
224
- "IVD5": {
225
- "Modic": datasets.Value(dtype="string"),
226
- "UP endplate": datasets.Value(dtype="string"),
227
- "LOW endplate": datasets.Value(dtype="string"),
228
- "Spondylolisthesis": datasets.Value(dtype="string"),
229
- "Disc herniation": datasets.Value(dtype="string"),
230
- "Disc narrowing": datasets.Value(dtype="string"),
231
- "Disc bulging": datasets.Value(dtype="string"),
232
- "Pfirrman grade": datasets.Value(dtype="string"),
233
- },
234
- "IVD6": {
235
- "Modic": datasets.Value(dtype="string"),
236
- "UP endplate": datasets.Value(dtype="string"),
237
- "LOW endplate": datasets.Value(dtype="string"),
238
- "Spondylolisthesis": datasets.Value(dtype="string"),
239
- "Disc herniation": datasets.Value(dtype="string"),
240
- "Disc narrowing": datasets.Value(dtype="string"),
241
- "Disc bulging": datasets.Value(dtype="string"),
242
- "Pfirrman grade": datasets.Value(dtype="string"),
243
- },
244
- "IVD7": {
245
- "Modic": datasets.Value(dtype="string"),
246
- "UP endplate": datasets.Value(dtype="string"),
247
- "LOW endplate": datasets.Value(dtype="string"),
248
- "Spondylolisthesis": datasets.Value(dtype="string"),
249
- "Disc herniation": datasets.Value(dtype="string"),
250
- "Disc narrowing": datasets.Value(dtype="string"),
251
- "Disc bulging": datasets.Value(dtype="string"),
252
- "Pfirrman grade": datasets.Value(dtype="string"),
253
- },
254
- "IVD8": {
255
- "Modic": datasets.Value(dtype="string"),
256
- "UP endplate": datasets.Value(dtype="string"),
257
- "LOW endplate": datasets.Value(dtype="string"),
258
- "Spondylolisthesis": datasets.Value(dtype="string"),
259
- "Disc herniation": datasets.Value(dtype="string"),
260
- "Disc narrowing": datasets.Value(dtype="string"),
261
- "Disc bulging": datasets.Value(dtype="string"),
262
- "Pfirrman grade": datasets.Value(dtype="string"),
263
- },
264
- "IVD9": {
265
- "Modic": datasets.Value(dtype="string"),
266
- "UP endplate": datasets.Value(dtype="string"),
267
- "LOW endplate": datasets.Value(dtype="string"),
268
- "Spondylolisthesis": datasets.Value(dtype="string"),
269
- "Disc herniation": datasets.Value(dtype="string"),
270
- "Disc narrowing": datasets.Value(dtype="string"),
271
- "Disc bulging": datasets.Value(dtype="string"),
272
- "Pfirrman grade": datasets.Value(dtype="string"),
273
- },
274
  }
275
  })
276
 
@@ -444,10 +364,10 @@ class SPIDER(datasets.GeneratorBasedBuilder):
444
  patient_grades = [
445
  x for x in grades_data if x['Patient'] == str(patient_id)
446
  ]
447
- # Pad radiological gradings so that data for all patients have
448
- # the same dimensions
449
- if len(patient_grades) < MAX_IVD:
450
- for i in range(len(patient_grades) + 1, MAX_IVD + 1):
451
  patient_grades.append({
452
  "Patient": f"{patient_id}",
453
  "IVD label": f"{i}",
@@ -460,9 +380,19 @@ class SPIDER(datasets.GeneratorBasedBuilder):
460
  "Disc bulging": "",
461
  "Pfirrman grade": "",
462
  })
463
- assert len(patient_grades) == MAX_IVD, "Rad. gradings not padded correctly"
464
- grades_dict[str(patient_id)] = patient_grades
465
- assert all([len(x) == MAX_IVD for x in grades_dict.values()])
 
 
 
 
 
 
 
 
 
 
466
 
467
  # Import image and mask data
468
  image_files = [
@@ -534,14 +464,7 @@ class SPIDER(datasets.GeneratorBasedBuilder):
534
  image_overview = overview_dict[scan_id]
535
 
536
  # Extract patient radiological gradings corresponding to patient
537
- patient_grades_dict = {}
538
- for item in grades_dict[patient_id]:
539
- key = f'IVD{item["IVD label"]}'
540
- value = {
541
- k:v for k,v in item.items()
542
- if k not in ['Patient', 'IVD label']
543
- }
544
- patient_grades_dict[key] = value
545
 
546
  # Prepare example return dict
547
  return_dict = {'patient_id':patient_id, 'scan_type':scan_type}
@@ -555,4 +478,4 @@ class SPIDER(datasets.GeneratorBasedBuilder):
555
  return_dict['rad_gradings'] = patient_grades_dict
556
 
557
  # Yield example
558
- yield (scan_id, return_dict)
 
37
 
38
  # Define constants
39
  N_PATIENTS = 257
40
+ MIN_IVD = 0
41
  MAX_IVD = 9
42
 
43
  # TODO: Add BibTeX citation
 
182
  "WindowWidth": datasets.Value(dtype="string"),
183
  },
184
  "rad_gradings": {
185
+ "IVD label": datasets.Sequence(datasets.Value("string")),
186
+ "Modic": datasets.Sequence(datasets.Value("string")),
187
+ "UP endplate": datasets.Sequence(datasets.Value("string")),
188
+ "LOW endplate": datasets.Sequence(datasets.Value("string")),
189
+ "Spondylolisthesis": datasets.Sequence(datasets.Value("string")),
190
+ "Disc herniation": datasets.Sequence(datasets.Value("string")),
191
+ "Disc narrowing": datasets.Sequence(datasets.Value("string")),
192
+ "Disc bulging": datasets.Sequence(datasets.Value("string")),
193
+ "Pfirrman grade": datasets.Sequence(datasets.Value("string")),
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
194
  }
195
  })
196
 
 
364
  patient_grades = [
365
  x for x in grades_data if x['Patient'] == str(patient_id)
366
  ]
367
+ # Pad so that all patients have same number of IVD observations
368
+ IVD_values = [x['IVD label'] for x in patient_grades]
369
+ for i in range(MIN_IVD, MAX_IVD + 1):
370
+ if str(i) not in IVD_values:
371
  patient_grades.append({
372
  "Patient": f"{patient_id}",
373
  "IVD label": f"{i}",
 
380
  "Disc bulging": "",
381
  "Pfirrman grade": "",
382
  })
383
+ assert len(patient_grades) == (MAX_IVD - MIN_IVD + 1), "Radiological\
384
+ gradings not padded correctly"
385
+
386
+ # Convert to sequences
387
+ df = (
388
+ pd.DataFrame(patient_grades)
389
+ .sort_values("IVD label")
390
+ .reset_index(drop=True)
391
+ )
392
+ grades_dict[str(patient_id)] = {
393
+ col:df[col].tolist() for col in df.columns
394
+ if col not in ['Patient']
395
+ }
396
 
397
  # Import image and mask data
398
  image_files = [
 
464
  image_overview = overview_dict[scan_id]
465
 
466
  # Extract patient radiological gradings corresponding to patient
467
+ patient_grades_dict = grades_dict[patient_id]
 
 
 
 
 
 
 
468
 
469
  # Prepare example return dict
470
  return_dict = {'patient_id':patient_id, 'scan_type':scan_type}
 
478
  return_dict['rad_gradings'] = patient_grades_dict
479
 
480
  # Yield example
481
+ yield scan_id, return_dict