Task 1: circle of Willis classification
The algorithm output should contain two classes per image: one for the anterior circle of Willis variant and one for the posterior circle of Willis variant.Classification algorithm output: a JSON file “result_Lippert.json” written into /output, containing a dictionary with the anterior and posterior class variant. See an example below.
{'Anterior class': 'g', 'Posterior class': 'h'}/result_Lippert.json
Task 2: circle of Willis quantification
Quantification algorithm output: two JSON files “result_diameters.json” and “result_angles.json” written into /output, containing dictionaries using the same keys as described in the annotator instructions. See an example below.
{'A2(L)': 2.228, 'A1(L)': 1.936, 'M1(L)': 2.183, 'ICA(L)': nan, 'Pcom(L)': nan, 'P1(L)': 2.163, 'BA': 2.832, 'VA(L)': 2.39, 'VA(R)': 2.931, 'P1(R)': 2.386, 'Pcom(R)': nan, 'ICA(R)': 2.924, 'M1(R)': 2.528, 'A1(R)': 1.916, 'A2(R)': nan}
/results_diameters.json
{‘A2_A1(L)’: 126.0, ‘ICA top(L)’: 99.0, ‘Pcom_ICA(L)’: nan, ‘M2_M2(L)’: nan, ‘BA top’: 96.0,
‘A2_A1(R)’: 135.0, ‘ICA top(R)’: 115.0, ‘Pcom_ICA(R)’: nan, ‘M2_M2(R)’: 86.0, ‘VBJ’: 113.0}
/results_angles.json