[v2] Added Celiac Disease use case

Added Celiac Disease use case

compute_stats.py

@@ -4,8 +4,8 @@
 import numpy as np
 
 parser = argparse.ArgumentParser()
-parser.add_argument('--outputs', default='./outputs/concepts/refined/colon/*.json', type=str, help='SKET results file.')
-parser.add_argument('--use_case', default='colon', choices=['colon', 'cervix', 'lung'], help='Considered use-case.')
+parser.add_argument('--outputs', default='./outputs/concepts/refined/aoec/colon/*.json', type=str, help='SKET results file.')
+parser.add_argument('--use_case', default='colon', choices=['colon', 'cervix', 'lung', 'celiac'], help='Considered use-case.')
 args = parser.parse_args()
 
 

dataset/original/celiac/README.md

@@ -0,0 +1,3 @@
+# Celiac datasets
+
+Put here datasets containing Celiac Disease pathology reports.

evaluate_sket.py

@@ -8,9 +8,9 @@
 
 
 parser = argparse.ArgumentParser()
-parser.add_argument('--gt', default='./ground_truth/lung/aoec/lung_labels_allDS.json', type=str, help='Ground truth file.')
-parser.add_argument('--outputs', default='./outputs/labels/aoec/lung/*.json', type=str, help='SKET results file.')
-parser.add_argument('--use_case', default='lung', choices=['colon', 'cervix', 'lung'], help='Considered use-case.')
+parser.add_argument('--gt', default='./ground_truth/celiac/aoec/celiac_labels_allDS.json', type=str, help='Ground truth file.')
+parser.add_argument('--outputs', default='./outputs/labels/aoec/celiac/*.json', type=str, help='SKET results file.')
+parser.add_argument('--use_case', default='celiac', choices=['colon', 'cervix', 'lung', 'celiac'], help='Considered use-case.')
 parser.add_argument('--hospital', default='aoec', choices=['aoec', 'radboud'], help='Considered hospital.')
 parser.add_argument('--debug', default=False, action='store_true', help='Whether to use evaluation for debugging purposes.')
 args = parser.parse_args()
@@ -41,6 +41,11 @@
         'Cancer - non-small cell cancer, squamous cell carcinoma': 'cancer_nscc_squamous',
         'Cancer - small cell cancer': 'cancer_scc',
         'Cancer - non-small cell cancer, large cell carcinoma': 'cancer_nscc_large'
+    },
+    'celiac': {
+        'Normal': 'normal',
+        'Celiac disease': 'celiac_disease',
+        'Non-specific duodenitis': 'duodenitis',
     }
 }
 
@@ -56,22 +61,25 @@ def main():
 
     gt = {}
     # prepare ground-truth for evaluation
-    if args.use_case == 'colon' and args.hospital == 'aoec':
-        gt = ground_truth
+    if args.hospital == 'aoec' or args.use_case == 'celiac':
+        ground_truth = ground_truth['groundtruths']
     else:
         ground_truth = ground_truth['ground_truth']
-        for data in ground_truth:
+    for data in ground_truth:
+        if args.hospital == 'aoec' or args.use_case == 'celiac':
+            rid = data['id_report']
+        else:
             rid = data['report_id_not_hashed']
 
-            if len(rid.split('_')) == 3 and args.hospital == 'aoec':  # contains codeint info not present within new processed reports
-                rid = rid.split('_')
-                rid = rid[0] + '_' + rid[2]
+        if len(rid.split('_')) == 3 and args.hospital == 'aoec':  # contains codeint info not present within new processed reports
+            rid = rid.split('_')
+            rid = rid[0] + '_' + rid[2]
 
-            gt[rid] = {label2class[args.use_case][label]: 0 for label in label2class[args.use_case].keys()}
-            for datum in data['labels']:
-                label = label2class[args.use_case][datum['label']]
-                if label in gt[rid]:
-                    gt[rid][label] = 1
+        gt[rid] = {label2class[args.use_case][label]: 0 for label in label2class[args.use_case].keys()}
+        for datum in data['labels']:
+            label = label2class[args.use_case][datum['label']]
+            if label in gt[rid]:
+                gt[rid][label] = 1
     # gt name
     gt_name = args.gt.split('/')[-1].split('.')[0]
 
@@ -93,14 +101,16 @@ def main():
     for rid, rdata in rs.items():
         if args.use_case == 'colon' and args.hospital == 'aoec' and '2ndDS' in args.gt:
             rid = rid.split('_')[0]
-        if args.hospital == 'radboud' and args.use_case == 'colon':
+        if args.hospital == 'radboud':
             sket[rid] = rdata['labels']
         else:
             sket[rid] = rdata
 
     # fix class order to avoid inconsistencies
     rids = list(sket.keys())
     classes = list(sket[rids[0]].keys())
+    if args.use_case == 'celiac':
+        classes.remove('inconclusive')
 
     # obtain ground-truth and SKET scores
     gt_scores = []
@@ -118,6 +128,7 @@ def main():
         if args.debug:
             first = True
         for c in classes:
+            #if c != 'inconclusive':
             gt_rscores.append(gt[rid][c])
             sket_rscores.append(sket[rid][c])
             if args.debug:  # perform debugging

run_med_sket.py

@@ -6,20 +6,21 @@
 warnings.filterwarnings("ignore", message=r"\[W008\]", category=UserWarning)
 
 parser = argparse.ArgumentParser()
-parser.add_argument('--src_lang', default='it', type=str, help='Considered source language.')
-parser.add_argument('--use_case', default='colon', choices=['colon', 'cervix', 'lung'], help='Considered use-case.')
+parser.add_argument('--src_lang', default='en', type=str, help='Considered source language.')
+parser.add_argument('--use_case', default='celiac', choices=['colon', 'cervix', 'lung', 'celiac'], help='Considered use-case.')
 parser.add_argument('--spacy_model', default='en_core_sci_sm', type=str, help='Considered NLP spacy model.')
-parser.add_argument('--w2v_model', default=False, action='store_true', help='Considered word2vec model.')
+parser.add_argument('--w2v_model', default=True, action='store_true', help='Considered word2vec model.')
 parser.add_argument('--fasttext_model', default=None, type=str, help='File path for FastText model.')
 parser.add_argument('--bert_model', default=None, type=str, help='Considered BERT model.')
-parser.add_argument('--string_model', default=False, action='store_true', help='Considered string matching model.')
+parser.add_argument('--string_model', default=True, action='store_true', help='Considered string matching model.')
 parser.add_argument('--gpu', default=None, type=int, help='Considered GPU device. If not specified (default to None), use CPU instead.')
-parser.add_argument('--thr', default=0.9, type=float, help='Similarity threshold.')
-parser.add_argument('--store', default=False, action='store_true', help='Whether to store concepts, labels, and graphs.')
+parser.add_argument('--thr', default=1.8, type=float, help='Similarity threshold.')
+parser.add_argument('--store', default=True, action='store_true', help='Whether to store concepts, labels, and graphs.')
 parser.add_argument('--rdf_format', default='all', choices=['n3', 'trig', 'turtle', 'all'], help='Whether to specify the rdf format for graph serialization. If "all" is specified, serialize w/ the three different formats')
 parser.add_argument('--raw', default=False, action='store_true', help='Whether to consider full pipeline or not.')
 parser.add_argument('--debug', default=False, action='store_true', help='Whether to use flags for debugging.')
-parser.add_argument('--dataset', default='', type=str, help='Dataset file path.')
+parser.add_argument('--preprocess', default=True, action='store_true', help='Whether to preprocess input data or not.')
+parser.add_argument('--dataset', default=None, type=str, help='Dataset file path.')
 args = parser.parse_args()
 
 
@@ -31,14 +32,14 @@ def main():
         dataset = args.dataset
     else:  # use sample "stream" dataset
         dataset = {
-                    'text': 'adenocarcinoma con displasia lieve, focalmente severa. Risultati ottenuti con biopsia al colon.',
-                    'gender': 'M',
+            "text": "polyp 40 cm: tubular adenoma with moderate dysplasia.",
+            'gender': 'F',
                     'age': 56,
-                    'id': 'test'
+                    'id': 'test_colon'
         }
 
     # use SKET pipeline to extract concepts, labels, and graphs from dataset
-    sket.med_pipeline(dataset, args.src_lang, args.use_case, args.thr, args.store, args.rdf_format, args.raw, args.debug)
+    sket.med_pipeline(dataset, args.preprocess, args.src_lang, args.use_case, args.thr, args.store, args.rdf_format, args.raw, args.debug)
 
     if args.raw:
         print('processed data up to concepts.')

run_sket.py

@@ -6,19 +6,19 @@
 warnings.filterwarnings("ignore", message=r"\[W008\]", category=UserWarning)
 
 parser = argparse.ArgumentParser()
-parser.add_argument('--dataset', default='./dataset/original/colon/radboud/ExaMode_3rdDS_Radboudumc_Colon_2ndBatch.xlsx', type=str, help='Dataset file.')
-parser.add_argument('--sheet', default='Sheet1', type=str, help='Considered dataset sheet.')
+parser.add_argument('--dataset', default='./dataset/original/celiac/aoec/ExaMode_0thDS_AOEC_Celiac.xlsx', type=str, help='Dataset file.')
+parser.add_argument('--sheet', default='Sheet 1', type=str, help='Considered dataset sheet.')
 parser.add_argument('--header', default=0, type=str, help='Header row within dataset.')
-parser.add_argument('--ver', default=1, type=str, help='Considered versioning for operations.')
-parser.add_argument('--use_case', default='colon', choices=['colon', 'cervix', 'lung'], help='Considered use-case.')
-parser.add_argument('--hospital', default='radboud', choices=['aoec', 'radboud'], help='Considered hospital.')
+parser.add_argument('--ver', default=2, type=str, help='Considered versioning for operations.')
+parser.add_argument('--use_case', default='celiac', choices=['colon', 'cervix', 'lung', 'celiac'], help='Considered use-case.')
+parser.add_argument('--hospital', default='aoec', choices=['aoec', 'radboud'], help='Considered hospital.')
 parser.add_argument('--spacy_model', default='en_core_sci_sm', type=str, help='Considered NLP spacy model.')
-parser.add_argument('--w2v_model', default=False, action='store_true', help='Considered word2vec model.')
+parser.add_argument('--w2v_model', default=True, action='store_true', help='Considered word2vec model.')
 parser.add_argument('--fasttext_model', default=None, type=str, help='File path for FastText model.')
 parser.add_argument('--bert_model', default=None, type=str, help='Considered BERT model.')
-parser.add_argument('--string_model', default=False, action='store_true', help='Considered string matching model.')
+parser.add_argument('--string_model', default=True, action='store_true', help='Considered string matching model.')
 parser.add_argument('--gpu', default=None, type=int, help='Considered GPU device. If not specified (default to None), use CPU instead.')
-parser.add_argument('--thr', default=2.0, type=float, help='Similarity threshold.')
+parser.add_argument('--thr', default=1.8, type=float, help='Similarity threshold.')
 parser.add_argument('--raw', default=False, action='store_true', help='Whether to return concepts within semantic areas (deployment) or mentions+concepts (debugging)')
 parser.add_argument('--debug', default=False, action='store_true', help='Whether to use flags for debugging.')
 args = parser.parse_args()