Getting Started with Medical Image Processing

Getting Started with Medical Image Processing in Python


Analyzing medical image in python is also possible. The source of medical image can come from X-Rays, MRI, CT and other.  X-Ray is the oldest medical imaging technique used to examine most area of the body to detect bone fracture and breaks. MRI stands for magnetic resonance imaging which produce slice images of the inside of the body like brain, cord, bones etc… CT stands for computational tomography or CAT computational axial tomography can scan the body and produce  detailed images of  internal organs, blood vessels and bones. 

To be able to open a saved MRI image, X-Ray image or CT image, you might need some procedure because they are binary images which means is not possible to open them as text directly. When a medical image scanner scans any organ of the body can send it to the workstation and produce a dcm image format  The format of medical image is mostly in dcm format. To open it in python, there is a special library for it, pydicom.

         DICOM stands for D-igital I-maging and CO-mmunications in M-edicine. You can install it in python by pip install pydicom.

 

We will show here the basic use of pydicom. I have a specific medical image (I can know the type of this image later on in this code) in my folder

# Reading medical image
>>> import pydicom
>>> image = pydicom.read_file('image-00004.dcm')
>>> image
(0008, 0005) Specific Character Set              CS: 'ISO_IR 100'
(0008, 0008) Image Type                          CS: ['ORIGINAL',
'PRIMARY', 'AXIAL', 'HELIX']
(0008, 0012) Instance Creation Date              DA: '20061012'
(0008, 0013) Instance Creation Time              TM: '091607.000000'
(0008, 0016) SOP Class UID                       UI: CT Image Storage
(0008, 0018) SOP Instance UID                    UI:
1.2.826.0.1.3680043.8.1055.1.20111102150759059.64998534.33344836
(0008, 0020) Study Date         
                DA: '20061012'
(0008, 0022) Acquisition Date                    DA: '20061012'
(0008, 0023) Content Date                        DA: '20061012'
(0008, 0030) Study Time                          TM: '090258.000000'
(0008, 0032) Acquisition Time                    TM: '085229.000000'
(0008, 0033) Content Time                        TM: '085229.844000'
(0008, 0060) Modality                            CS: 'CT'
(0008, 1030) Study Description                   LO: 'CT1 abdomen'
(0008, 1032)  Procedure Code
Sequence   1 item(s) ---- 
   (0008, 0100) Code
Value                          SH:
'CTABDOM'
   (0008, 0102) Coding Scheme
Designator            SH: 'XPLORE'
   (0008, 0104) Code
Meaning                        LO: 'CT1
abdomen'
   ---------
(0008, 103e) Series Description                  LO: 'ARTERIELLE'
(0008, 1111)  Referenced
Performed Procedure Step Sequence   1
item(s) ---- 
   (0008, 1150) Referenced
SOP Class UID            UI: Modality
Performed Procedure Step SOP Class
   (0008, 1155) Referenced
SOP Instance UID         UI:
1.2.840.113704.1.111.5104.1160636572.51
   ---------
(0010, 0010) Patient's Name                      PN: 'Anonymized'
(0010, 0020) Patient ID                          LO: '0'
(0010, 1010) Patient's Age                       AS: '000Y'
(0018, 0010) Contrast/Bolus Agent                LO: 'CONTRAST'
(0018, 0022) Scan Options                        CS: 'HELIX'
(0018, 0050) Slice Thickness                     DS: "1.0"
(0018, 0060) KVP                                 DS: "120.0"
(0018, 0088) Spacing Between Slices              DS: "0.5"
(0018, 0090) Data Collection Diameter            DS: "302.0"
(0018, 1030) Protocol Name                       LO: 'ART.RENALES
12/Abdomen/Hx'
(0018, 1100) Reconstruction Diameter             DS: "302.0"
(0018, 1120) Gantry/Detector Tilt                DS: "0.0"
(0018, 1130) Table Height                        DS: "151.0"
(0018, 1140) Rotation Direction                  CS: 'CW'
(0018, 1151) X-Ray Tube Current                  IS: "400"
(0018, 1152) Exposure                            IS: "300"
(0018, 1160) Filter Type                         SH: 'B'
(0018, 1210) Convolution Kernel                  SH: 'B'
(0018, 5100) Patient Position                    CS: 'FFS'
(0020, 000d) Study Instance UID                  UI:
1.2.826.0.1.3680043.8.1055.1.20111102150758591.92402465.76095170
(0020, 000e) Series Instance UID                 UI:
1.2.826.0.1.3680043.8.1055.1.20111102150758591.96842950.07877442
(0020, 0011) Series Number                       IS: "6168"
(0020, 0013) Instance Number                     IS: "5"
(0020, 0032) Image Position (Patient)            DS: [-151.493508, -36.6564417,
1293]
(0020, 0037) Image Orientation (Patient)         DS: [1, 0, 0, 0, 1, 0]
(0020, 0052) Frame of Reference UID              UI:
1.2.840.113704.1.111.3704.1160637109.3
(0020, 1041) Slice Location                      DS: "-323.0"
(0020, 4000) Image Comments                      LT: 'JPEG 2000 lossless -
Version 4.0.2 (c) Image Devices GmbH'
(0028, 0002) Samples per Pixel                   US: 1
(0028, 0004) Photometric Interpretation          CS: 'MONOCHROME2'
(0028, 0010) Rows                                US: 512
(0028, 0011) Columns                             US: 512
(0028, 0030) Pixel Spacing                       DS: [0.58984375, 0.58984375]
(0028, 0100) Bits Allocated                      US: 16
(0028, 0101) Bits Stored                         US: 12
(0028, 0102) High Bit                            US: 11
(0028, 0103) Pixel Representation                US: 0
(0028, 1050) Window Center                       DS: [00050, 00050]
(0028, 1051) Window Width                        DS: [00350, 00350]
(0028, 1052) Rescale Intercept                   DS: "-1000.0"
(0028, 1053) Rescale Slope                       DS: "1.0"
(0028, 2110) Lossy Image Compression             CS: '01'
(0028, 2112) Lossy Image Compression Ratio       DS: "5.881756"
(0040, 0007) Scheduled Procedure Step Descriptio LO: 'CT1 abdomen'
(0040, 0008)  Scheduled
Protocol Code Sequence   1 item(s) ---- 
   (0008, 0100) Code
Value                          SH:
'CTABDOM'
   (0008, 0102) Coding Scheme
Designator            SH: 'XPLORE'
   (0008, 0104) Code
Meaning                        LO: 'CT1
abdomen'
   ---------
(0040, 0009) Scheduled Procedure Step ID         SH: 'A10026177758'
(0040, 0254) Performed Procedure Step Descriptio LO: 'CT1 abdomen'
(0040, 0260)  Performed
Protocol Code Sequence   1 item(s) ---- 
   (0008, 0100) Code
Value                          SH:
'CTABDOM'
   (0008, 0102) Coding Scheme
Designator            SH: 'XPLORE'
   (0008, 0104) Code
Meaning                        LO: 'CT1
abdomen'
   ---------
(0040, 0275)  Request
Attributes Sequence   1 item(s) ---- 
   (0040, 0007) Scheduled
Procedure Step Descriptio LO: 'CT1 abdomen'
   (0040, 0008)  Scheduled Protocol Code Sequence   1 item(s) ---- 
      (0008, 0100) Code
Value                          SH:
'CTABDOM'
      (0008, 0102) Coding
Scheme Designator            SH: 'XPLORE'
      (0008, 0104) Code
Meaning                        LO: 'CT1
abdomen'
      ---------
   (0040, 0009) Scheduled
Procedure Step ID         SH:
'A10026177758'
   (0040, 1001) Requested
Procedure ID              SH:
'A10026177757'
   ---------
(0040, 1001) Requested Procedure ID              SH: 'A10026177757'
(7fe0, 0010) Pixel Data                          OW: Array of 89158
elements

 

I decided to print all the details of image to give you a full data that a medical image can contain

 

>>> len(image)

68

>>> # Checking patient name

>>> image.PatientName

'Anonymized'

It’s important to keep the Patient Name in anaonym for the ethical reason.

>>> image.PatientPosition  

'FFS'

FFS stands for Feet First-Supine.

>>> image.Modality

'CT'

This medical image has been token via CT

>>> image.StudyDate

'20061012'

>>> image.PixelSpacing

[0.58984375, 0.58984375]

>>> image.SliceLocation

"-323.0"

>>> image.XRayTubeCurrent

"400"

To go forward and know more things you can use in pydicom

>>> dir(image)  # You will have a lot of function that you can use to output a lot of result from the medical image CT

 

In the next tutorial, we will try to work more with medical image analyzing by extracting some data and visualizing it with matplotlib


Comments

Post a Comment

Popular posts from this blog

Farming land size owned by households per region in Senegal

Image
   Average farm size per household and region (ha) in Senegal 2001-2016   Agriculture is the best factor of the growing sector in Senegal. Almost all the region  except Dakar lives in farming.  We will analyze the data of "Farm size by household and region" from (DAPSA: Direction de l’Analyse, de la Prévision et des Statistiques Agricoles - Office of Agricultural Analysis, Forecasting and Statistics in Senegal). You can access to the data Let's Explore the Data: The head of our data and the tail (last rows) Senegal counts 14 regions. Let’s check the shape of our data to verify the presence of all regions. To do so, I use python pandas library to prove that. We could use eye verification since our data is not large.  In the 14 regions of Senegal, which region has the highest farming ground per household in every year? The best method is to sort the values  Our bar plot graph is representative and shows us that ‘Kaffrine’ is the highest region with farmi...
Read more