data_preparation.py

train_path = '../data/train/'


def load_one_stack(case, data_path=train_path, plane='coronal'):
    fpath = '{}/{}/{}.npy'.format(data_path, plane, case)
    return np.load(fpath)

def load_stacks(case, data_path=train_path):
    x = {}
    planes = ['coronal', 'sagittal', 'axial']
    for i, plane in enumerate(planes):
        x[plane] = load_one_stack(case, plane=plane)
    return x

def load_cases(train=True, n=None):
    assert (type(n) == int) and (n < 1250)
    if train:
        case_list = pd.read_csv('../data/train-acl.csv', names=['case', 'label'], header=None,
                               dtype={'case': str, 'label': np.int64})['case'].tolist()        
    else:
        case_list = pd.read_csv('../data/valid-acl.csv', names=['case', 'label'], header=None,
                               dtype={'case': str, 'label': np.int64})['case'].tolist()        
    cases = {}
    
    if n is not None:
        case_list = case_list[:n]
        
    for case in tqdm_notebook(case_list, leave=False):
        x = load_stacks(case)
        cases[case] = x
    return cases

cases = load_cases(n=100)