ValueError: total size of new array must be unchanged

[Andrew Hoole]

Hi 

I'm trying to write a simple script to convert a DICOM SC file into a JPEG image. It works when I run it the first time with one image but as soon as I try another image which has a slightly different size I get the following error:

Traceback (most recent call last):

  File "dcmjpg.py", line 43, in <module>

    plt.imshow(data.pixel_array)

  File "C:\Anaconda\lib\site-packages\pydicom-1.0.0a1-py2.7.egg\pydicom\dataset.

py", line 678, in pixel_array

    return self._get_pixel_array()

  File "C:\Anaconda\lib\site-packages\pydicom-1.0.0a1-py2.7.egg\pydicom\dataset.

py", line 670, in _get_pixel_array

    self._pixel_array = self._pixel_data_numpy()

  File "C:\Anaconda\lib\site-packages\pydicom-1.0.0a1-py2.7.egg\pydicom\dataset.

py", line 509, in _pixel_data_numpy

    pixel_array = pixel_array.reshape(self.Rows, self.Columns, self.SamplesPerPi

xel)

ValueError: total size of new array must be unchanged

If I run it again using the original image it works fine again. I am running it each time from the command line as:

         python dcmjpg.py inputfile outputdir

I have seen this behavior on both my Windows 7 64 bit machine and an old XP 32bit machine both of which have an Anaconda 2.7 distro of Python installed on them. 

Searching on google I found this link (https://travis-ci.org/darcymason/pydicom/jobs/141981764) which seems to highlight the same problem. 

The code itself up until the point which throws the error (line plt.imshow(data.pixel_array)) is very simple see below:

# -*- coding: utf-8 -*-

"""

Title: DcmJpg.py 

Function: Converts a DICOM SC file from Prosoma to a JPEG image file with some 

          annotation added to the image to aid in the identification of the 

          image and the appropriate association of the image. 

          

Usage: dcmjpg input_dicom_file output_dir

Written by: A. Hoole

Date: 14/9/2016

Version: 0.9

"""

import sys

import pydicom as dicom 

import matplotlib.pyplot as plt

# Get the input filename from the command line and the output directory

dcmfile = sys.argv[1]

outdir = sys.argv[2]

# Read in the dicom file 

data = dicom.read_file(dcmfile, force=True)

# Add in the PlanarConfiguration tag to the DICOM, if left out it should be 

# assumed to be interlaced but imshow fails unless it is explicitly defined

data.add_new([0x0028,0x0006],'US',0)

# Create a figure based on the ColxRow of the DICOM data

XSize = data.Columns/100.

YSize = data.Rows/100.

fig = plt.figure(figsize=(XSize,YSize),frameon=False,dpi=100)

# Set the plot area axes so that the image fills the whole figure with no border

fig.add_axes([0,0,1,1])

# Remove any axis markers from the image

plt.xticks(())

plt.yticks(())

# Display the pixel data of the image

plt.imshow(data.pixel_array)

Any ideas what might be causing this problem? It almost seems as if there is some residual memory of the pixel_array size from one runtime to another runtime which is odd. Maybe I should be clearing something out or initializing something at the start to prevent this problem happening. 

Any ideas would be most appreciated. 

Regards

Andrew Hoole

[Darcy Mason]

I just had a quick read, but sounds the same as https://github.com/darcymason/pydicom/pull/271.  I believe there is a merge done with code there, and then again a further fix in PR#274. Perhaps see if that last commit solves your problem.

-Darcy

[Andrew Hoole]

I'll give that a go. Do I just need to download and upgrade my installation of pydicom to the last commit. 

Regards

Andrew

[Darcy Mason]

It might depend somewhat on how you are set up - e.g. virtual environment, etc., but probably the easiest way in any case is to use pip with a specific commit specified. 

See, e.g. http://stackoverflow.com/questions/13685920/install-specific-git-commit-with-pip

Again, depending on your setup, I suppose you might need the upgrade flag.  I'm not sure if that still works as in that link above.  If all else fails you could set up a virtual environment and install that specific commit in the virtual environment.

-Darcy