how to convert dicom to jpeg

[wanghaisheng]

what i want is to get a jpeg file for each dcm file ?is there any best practices of pydicom ,i find nothing looking over documentation

[scaramallion]

Not tested, but you should be able to use pillow and numpy:

from PIL.Image import fromarray

ds = read_file('file-in.dcm')
im = fromarray(ds.pixel_array)
im.save('file-out.jpg')

You will likely have to do some work to ensure you have the image data in a suitable format prior to saving.

[mrbean-bremen]

I think this can be closed as answered.

[tedivm]

This doesn't work for me, so I'm guessing it doesn't work for others. I'm including some code that converts a dicom file into a png for the next person who ends up here via google :-)

import numpy as np
import png
import pydicom

ds = pydicom.dcmread(path)

shape = ds.pixel_array.shape

# Convert to float to avoid overflow or underflow losses.
image_2d = ds.pixel_array.astype(float)

# Rescaling grey scale between 0-255
image_2d_scaled = (np.maximum(image_2d,0) / image_2d.max()) * 255.0

# Convert to uint
image_2d_scaled = np.uint8(image_2d_scaled)

# Write the PNG file
with open(destination, 'wb') as png_file:
    w = png.Writer(shape[1], shape[0], greyscale=True)
    w.write(png_file, image_2d_scaled)

[wanghaisheng]

@tedivm wonderful

[rohit-ptl]

Thanks @tedivm. Wrapped your code into a function for convenient copy-paste

import numpy as np
import png, os, pydicom

source_folder = r'path\to\source'
output_folder = r'path\to\output\folder'


def dicom2png(source_folder, output_folder):
    list_of_files = os.listdir(source_folder)
    for file in list_of_files:
        try:
            ds = pydicom.dcmread(os.path.join(source_folder,file))
            shape = ds.pixel_array.shape

            # Convert to float to avoid overflow or underflow losses.
            image_2d = ds.pixel_array.astype(float)

            # Rescaling grey scale between 0-255
            image_2d_scaled = (np.maximum(image_2d,0) / image_2d.max()) * 255.0

            # Convert to uint
            image_2d_scaled = np.uint8(image_2d_scaled)

            # Write the PNG file
            with open(os.path.join(output_folder,file)+'.png' , 'wb') as png_file:
                w = png.Writer(shape[1], shape[0], greyscale=True)
                w.write(png_file, image_2d_scaled)
        except:
            print('Could not convert: ', file)


dicom2png(source_folder, output_folder)

[tedivm]

@Rohit-Patel - you forgot to import os ;-)

[samra-irshad]

This doesn't work for me, so I'm guessing it doesn't work for others. I'm including some code that converts a dicom file into a png for the next person who ends up here via google :-)

import numpy as np
import png
import pydicom

ds = pydicom.dcmread(path)

shape = ds.pixel_array.shape

# Convert to float to avoid overflow or underflow losses.
image_2d = ds.pixel_array.astype(float)

# Rescaling grey scale between 0-255
image_2d_scaled = (np.maximum(image_2d,0) / image_2d.max()) * 255.0

# Convert to uint
image_2d_scaled = np.uint8(image_2d_scaled)

# Write the PNG file
with open(destination, 'wb') as png_file:
    w = png.Writer(shape[1], shape[0], greyscale=True)
    w.write(png_file, image_2d_scaled)

Only solution that worked for my problem, thank you!

[kyshel]

This doesn't work for me, so I'm guessing it doesn't work for others. I'm including some code that converts a dicom file into a png for the next person who ends up here via google :-)

import numpy as np
import png
import pydicom

ds = pydicom.dcmread(path)

shape = ds.pixel_array.shape

# Convert to float to avoid overflow or underflow losses.
image_2d = ds.pixel_array.astype(float)

# Rescaling grey scale between 0-255
image_2d_scaled = (np.maximum(image_2d,0) / image_2d.max()) * 255.0

# Convert to uint
image_2d_scaled = np.uint8(image_2d_scaled)

# Write the PNG file
with open(destination, 'wb') as png_file:
    w = png.Writer(shape[1], shape[0], greyscale=True)
    w.write(png_file, image_2d_scaled)

@tedivm
got an error， maybe it can be optimized?

Traceback (most recent call last):
  File "D:/ml/untitled/dcm2png.py", line 81, in <module>
    dcm2png(file,'png2\\'+file+'.png')
  File "D:/ml/untitled/dcm2png.py", line 16, in dcm2png
    shape = ds.pixel_array.shape
  File "D:\ProgramData\Anaconda3\envs\py36\lib\site-packages\pydicom\dataset.py", line 778, in __getattr__
    return object.__getattribute__(self, name)
  File "D:\ProgramData\Anaconda3\envs\py36\lib\site-packages\pydicom\dataset.py", line 1615, in pixel_array
    self.convert_pixel_data()
  File "D:\ProgramData\Anaconda3\envs\py36\lib\site-packages\pydicom\dataset.py", line 1324, in convert_pixel_data
    self._convert_pixel_data_without_handler()
  File "D:\ProgramData\Anaconda3\envs\py36\lib\site-packages\pydicom\dataset.py", line 1434, in _convert_pixel_data_without_handler
    raise last_exception
  File "D:\ProgramData\Anaconda3\envs\py36\lib\site-packages\pydicom\dataset.py", line 1414, in _convert_pixel_data_without_handler
    self._do_pixel_data_conversion(handler)
  File "D:\ProgramData\Anaconda3\envs\py36\lib\site-packages\pydicom\dataset.py", line 1441, in _do_pixel_data_conversion
    arr = handler.get_pixeldata(self)
  File "D:\ProgramData\Anaconda3\envs\py36\lib\site-packages\pydicom\pixel_data_handlers\numpy_handler.py", line 263, in get_pixeldata
    "Unable to convert the pixel data: one of Pixel Data, Float "
AttributeError: Unable to convert the pixel data: one of Pixel Data, Float Pixel Data or Double Float Pixel Data must be present in the dataset

[ykuo2]

I'm working on this problem recently. It turns out that there are many things to consider on converting DICOM to JPG.
For example, we have to check tags/attributes below and adjust accordingly

• window center(level) and window width
• does it need to apply VOI LUT function
• does it need negative inversion
• does it contains more than one images (eg. fluoroscopic images)
• does it contains filetype that is not an image (eg. PDF file)
• if it is a colored file, there are some matrix manipulation to work on
• and many amazing DICOM tags/attributes...

Here is my preliminary package , please have a look, hope it would help.

import dicom2jpg

dicom_img_01 = "/Users/user/Desktop/img01.dcm"
dicom_dir = "/Users/user/Desktop/Patient_01"
export_location = "/Users/user/Desktop/BMP_files"

# convert single DICOM file to jpg format
dicom2jpg.dicom2jpg(dicom_img_01)  

# convert all DICOM files in dicom_dir folder to png format
dicom2jpg.dicom2png(dicom_dir)  

# convert all DICOM files in dicom_dir folder to bmp, to a specified location
dicom2jpg.dicom2bmp(dicom_dir, target_root=export_location) 

[tommy-qichang]

I think there's some fundamental issues here. The Dicom image data has much larger range (eg. CT: -1024 HU to 3071 HU)and if you convert it to uint8, you loss information.

[kyshel]

@tommy-qichang so, is there some recommend ways to convert that not lost info? maybe convert to 3 channel ?

[tommy-qichang]

I think PIL support I16 so @scaramallion 's solution should be better.

[scaramallion]

PIL's JPEG2000 plugin doesn't support encoding signed data. It might "work" but it won't be correct.

[yehudahkay]

None of the answers worked for me with a single frame dicom, the colors were always off. Strange when using a multi frame dicom the following code worked fine

for index, slice in enumerate(pixel_array):
                cv2.imwrite(filename, slice)