--- a
+++ b/scripts/data/parse_mdai_annotations.py
@@ -0,0 +1,102 @@
+import argparse
+import json
+from tqdm import tqdm
+from datetime import datetime
+from collections import defaultdict
+import pdb
+import pandas as pd
+
+"""
+Script to parse JSON file exported from md.ai
+JSON structure is detailed in https://docs.md.ai/data/json/ 
+
+Output is either:
+    (1) a dict mapping image (SOPInstanceUID) to list of annotation data
+    OR
+    (2) dataset object JSON with annotation dict for each image series
+
+Notes:
+    - JSON: ['id', 'createdAt', 'updatedAt', 'name', 'description', 'isPrivate', 'users', 'labelGroups', 'datasets']
+    
+    - meta['labelGroups']: list of dicts with keys: 
+        ['id','createdAt', 'updatedAt', 'name', 'description', 'type', 
+            'labels': [{'id', 'parentId', 'createdAt', 'updatedAt', 'name', 'shortName', 
+            'description', 'color', 'type,', ' scope', 'annotationMode', 'radlexTagIds': []}] ]
+    
+    - meta['datasets']: list of dicts: ['id', 'type', 'createdAt', 'updatedAt', 'name', 'description', 'studies', 'annotations']
+    
+    - meta['datasets'][i]['annotations']: list of dicts for each annotation (image-level):
+        ['id', 'parentId', 'createdAt', 'createdById', 'updatedAt', 'updatedById', 'modelId', 'StudyInstanceUID', 'SeriesInstanceUID', 
+            'SOPInstanceUID', 'labelId', 'annotationNumber', 'height', 'width', 'data', 'note', 'radlexTagIds', 'isImported', 
+            'reviewsPositiveCount', 'reviewsNegativeCount']
+
+    - date format: %Y-%m-%dT%H:%M:%S.%fZ
+    - for box annotations: x,y are upper left corner 
+    - 'SOPInstanceUID': slice id (single image)
+    - 'SeriesInstanceUID': series id (single volume) 
+    - 'StudyInstanceUID': exam id (multiple volumes)
+    - 'height', 'width': image h, w (not annotation)
+"""
+
+INCLUDE_AFTER = datetime.strptime('10-08-2020', '%m-%d-%Y')  # annotations made before INCLUDE_AFTER were used to test md.ai
+ANNOT_COMMENTS = pd.read_csv('/Mounts/rbg-storage1/datasets/NLST/mdai/annotation_comments_12062020.csv')
+
+def scale_annotations(annotation, annotation_meta):
+    '''
+    Scale annotation (bounding boxes) to values in [0,1] by dividing by image height and width (annotation_meta)
+    '''
+    annotation['x'] /= annotation_meta['width']
+    annotation['width'] /= annotation_meta['width']
+    annotation['y'] /= annotation_meta['height']
+    annotation['height'] /= annotation_meta['height']
+    return annotation
+
+parser = argparse.ArgumentParser()
+parser.add_argument('--annotation_json_path', type = str, help = 'JSON exported from md.ai', default = '/Mounts/rbg-storage1/datasets/NLST/mdai/mdai_mit_project_poBGbqle_annotations_labelgroup_all_2020-11-25-030811.json')
+parser.add_argument('--output_json_path', type = str, help = 'Where to export parsed annotations data')
+parser.add_argument('--output_is_dataset_obj', action = 'store_true', default = False, help = 'Whether incorporating annotations into dataset json directly')
+
+if __name__ == "__main__":
+    args = parser.parse_args()
+    annotation_metadata_json = json.load(open(args.annotation_json_path, 'r'))
+    annotation_dict =  {}
+
+    USERS = { user['id']: user['name'] for user in annotation_metadata_json['users']  }
+    
+    for dataset_dict in annotation_metadata_json['datasets']:
+        for annotation_meta_dict in tqdm(dataset_dict['annotations']):
+            if annotation_meta_dict['data'] is None:
+                continue
+            date = datetime.strptime(annotation_meta_dict['createdAt'], '%Y-%m-%dT%H:%M:%S.%fZ')
+            if not(date > INCLUDE_AFTER):
+                continue
+            img_id = annotation_meta_dict['SOPInstanceUID']
+            series_id = annotation_meta_dict['SeriesInstanceUID']
+            exam_id = annotation_meta_dict['StudyInstanceUID']
+            if series_id not in annotation_dict:
+                annotation_dict[series_id] = defaultdict(list)
+            annotation_meta_dict['data'] = scale_annotations(annotation_meta_dict['data'], annotation_meta_dict)
+            annotation_meta_dict['data']['user'] = USERS[annotation_meta_dict['createdById']]
+            annotation_dict[series_id][img_id].append( annotation_meta_dict['data'] )
+     
+    for series_id in annotation_dict.keys():
+        if series_id in list(ANNOT_COMMENTS['series_uid']):
+            if 'FF' in str(ANNOT_COMMENTS[ANNOT_COMMENTS['series_uid'] == series_id]['comments for Peter']):
+                for img_id in annotation_dict[series_id].keys():
+                    annotation_dict[series_id][img_id] = [ i for i in annotation_dict[series_id][img_id]  if 'fintelmann' in i['user'] ]
+    
+    if not args.output_is_dataset_obj:
+        json.dump(annotation_dict, open(args.output_json_path, 'w'))
+    
+    else:
+        output_json = json.load(open(args.output_json_path, 'r')) 
+
+        for mrn_row in tqdm(output_json):
+            for exam_dict in mrn_row['accessions']:
+                for series_id, series_dict in exam_dict['image_series'].items():
+                    if series_id in annotation_dict:
+                        series_dict['annotations'] = annotation_dict[series_id]
+                    else:
+                        continue
+        
+        json.dump(output_json, open(args.output_json_path, 'r'))