(0x4465e80): Too many samples map outside moving image buffer: 784 / 4096

[black0017]

Hello,

What I have tried:

I am registering a dataset of CBCT's to CT. While I have 80 patients with multiple CBCT for each one, for two patients I get this error:

Error log:

itk::ExceptionObject (0x3d6c550)
Location: "ElastixTemplate - Run()" 
File: /home/utilisateur/build/Elastix/Common/CostFunctions/itkAdvancedImageToImageMetric.hxx
Line: 1007
Description: itk::ERROR: AdvancedMattesMutualInformationMetric(0x4465e80): **Too many samples map outside moving image buffer: 784 / 4096**


Error occurred during actual registration.

I resized the CT image to the CBCT so that they have the same spatial dimensions. It's very strange that this error shows up only for registration in the whole dataset

How can I possibly fix this issue?

Thanks

[velonica0]

How did you solve this problem, is it because the resolution of the two pictures is different?

[black0017]

Hi @velonica0 , to be honest I don't remember, it's been a long time, but image resolutions was not the problem. I guess it had to do with the "empty" slices on my CBCT image but i really cannot recall.

I used the following 3 functions please take a look and see if something is useful to you:

import os

import SimpleITK as sitk
import numpy as np


def register_and_save(ref_img, moving_CBCT_path, save_path, debug=False):
    mov_img = sitk.ReadImage(moving_CBCT_path)
    # print('cbct shape:',sitk.GetArrayFromImage(mov_img).shape)
    # print('ct shape:',sitk.GetArrayFromImage(ref_img).shape)

    resultImage = register(ref_img, mov_img, reg_type='translation',debug=debug)

    print('resulted img shape:', sitk.GetArrayFromImage(resultImage).shape)
    result_np = sitk.GetArrayFromImage(resultImage)
    #result_np = delete_empy_slices(result_np)
    resultImage = sitk.GetImageFromArray(result_np)

    save_nifty(resultImage, save_path, moving_CBCT_path)


def register(ref_img, mov_img, reg_type='translation', debug=False):
    # "affine" and 'translation' is supported
    parameterMap = sitk.GetDefaultParameterMap(reg_type)
    parameterMap['DefaultPixelValue'] = ['-1000']  # AIR in Housenfield units
    parameterMap['MaximumNumberOfIterations'] = ['512']  # more iteration for better convergence

    if debug:
        itk_filter = sitk.ElastixImageFilter()
        itk_filter.LogToConsoleOn()  #debug
        itk_filter.LogToFileOn()  #debug
        itk_filter.SetFixedImage(ref_img)
        itk_filter.SetMovingImage(mov_img)
        itk_filter.SetParameterMap(parameterMap)
        resultImage = itk_filter.Execute()
    else:
        resultImage = sitk.Elastix(ref_img, mov_img, parameterMap)
    return resultImage


def delete_empy_slices(result_np):
    slices = result_np.shape[0]
    mean_val_per_slice = np.mean(result_np, axis=1).mean(axis=1)
    # print(mean_val_per_slice.shape)
    non_empty_slices = []
    for i in range(slices):
        if mean_val_per_slice[i] >= -900:
            non_empty_slices.append(result_np[i, ...])
    print('len non empty slices:', len(non_empty_slices), '/',mean_val_per_slice.shape[0])
    cropped_img = np.asarray(non_empty_slices)
    return cropped_img

Best :)

[mathiser]

For anyone with the same issue - this error disappeared for me when increasing parameterMap['MaximumNumberOfIterations'] = ['512'] from the default