DICOM data (UltraSound) displayed in wrong color

[star50037]

It looks in dicom viewer

ds = pydicom.dcmread("USImg.DCM")
arr = ds.pixel_array
plt.imshow(arr)
plt.show()

But getting following result

color incorrect and has some stripe
i try use pydicom.pixel_data_handlers.util like apply_color_lut()
but almost no change in result like #1430

ds = pydicom.dcmread("USImg.DCM")
rgb = pydicom.pixel_data_handlers.util.apply_voi_lut(arr, ds,index= 0)
f, axarr = plt.subplots(1,2) 
axarr[0].imshow(arr)
axarr[1].imshow(rgb)
axarr[0].title.set_text('arr')
axarr[1].title.set_text('rgb')
plt.show()

Here are the group 0028 elements , dicom_file and my environment

python:'3.7.9'
pydicom:'2.3.0'
matplotlib : '3.5.1'

(0028, 0002) Samples per Pixel                   US: 3
(0028, 0004) Photometric Interpretation          CS: 'RGB'
(0028, 0006) Planar Configuration                US: 0
(0028, 0010) Rows                                US: 720
(0028, 0011) Columns                             US: 960
(0028, 0014) Ultrasound Color Data Present       US: 1
(0028, 0100) Bits Allocated                      US: 8
(0028, 0101) Bits Stored                         US: 8
(0028, 0102) High Bit                            US: 7
(0028, 0103) Pixel Representation                US: 0
(0028, 0301) Burned In Annotation                CS: 'YES'
(0028, 2110) Lossy Image Compression             CS: '00'

US_dcm.zip

[mrbean-bremen]

Your initial code snippet was correct - setting a VOI LUT on an RGB image does not make sense.
The problem seems to be the transfer syntax - Big Endian Explicit - which seems not to be correctly handled in this case. This transfer syntax is long deprecated, and I haven't seen it in the wild for many years now, but we still should support it.

If converting the image to Little Endian first, for example using dcmodify +te (from dcmtk), it will display correctly using your initial code. I'm not sure yet what the problem is here (pydicom should still support Big Endian data), but we will have a closer look at this.

As an aside: I first opened the file using the Weasis viewer, which is generally quite a good viewer, and it had the same output as matplotlib - support of Big Endian transfer syntax seems to be on the decline...

[mrbean-bremen]

I could reproduce the problem with a test dataset, documented in issue #1680.

[star50037]

The file has anonymized ,all about patient's data is fake