tio.HistogramStandardization.train() divide by zero RuntimeError

[GDSALex]

Hi, i was trying to implement the tio.HistogramStandardization function to my code to do a segmentation neural network, but following the torchio examples I keep getting the same error when training the HistogramStandardization function: RuntimeWarning: divide by zero encountered in divide slopes = (s2 - s1) / (pc2 - pc1). I think the rest of the code doesn't have any problems so i don't know why this error might be appearing. The files of landmarks_{modality}.npy are also empty after the code has been executed. Here is my code:

root_dir = 'C:/Users/Alejandro/Desktop/ASNR-MICCAI-BraTS2023-GLI-Challenge-TrainingData'
output_dir = 'C:/Users/Alejandro/Desktop/BraTS Challenge CODE'


def collect_image_paths(root_dir):
    volume_paths = {}
    
    def process_directory(dir_path):
        nonlocal volume_paths
        for entry in os.scandir(dir_path):
            if entry.is_dir():
                process_directory(entry.path)
            elif entry.is_file() and entry.name.endswith('.nii.gz'):
                parts = entry.name.split('-')
                if len(parts) >= 2:
                    modality = parts[-1].split('.')[0]
                    volume_paths.setdefault(modality, []).append(entry.path)
                else:
                    print(f"Ignoring file: {entry.name}. Filename format doesn't match the expected pattern.")

     
    process_directory(root_dir)
    
    return volume_paths

def process_volume_paths(modality, volume_paths, output_dir):
    landmarks = tio.HistogramStandardization.train(
        volume_paths,
        output_path=os.path.join(output_dir, f'landmarks_{modality}.pt')
    )
    np.save(os.path.join(output_dir, f'landmarks_{modality}.npy'), landmarks)
    return modality, landmarks




volume_paths_dict = collect_image_paths(root_dir)


landmarks_dict = {}
with concurrent.futures.ThreadPoolExecutor() as executor:
    future_to_modality = {
        executor.submit(process_volume_paths, modality, volume_paths, output_dir): modality
        for modality, volume_paths in volume_paths_dict.items()
    }
    for future in concurrent.futures.as_completed(future_to_modality):
        modality = future_to_modality[future]
        try:
            modality, landmarks = future.result()
            landmarks_dict[modality] = landmarks
        except Exception as exc:
            print(f"Exception processing {modality}: {exc}")


def transform():
    include = ['t1c', 't1n', 't2f', 't2w', 'seg']
    transform = tio.Compose([
        tio.ToCanonical(include=include),
        #tio.CropOrPad((128, 192, 128), include=include),
        tio.RandomFlip(axes=['LR', 'AP', 'IS'], p=0.25, include=include),
        tio.HistogramStandardization(landmarks_dict, include=include),
        tio.ZNormalization(include=include),
        tio.OneOf({
            tio.RandomAffine(include=include): 0.7,
            tio.RandomElasticDeformation(include=include): 0.3,
        }),
        tio.RandomNoise(p=0.2, include=include)
    ])
    return transform