Anda di halaman 1dari 19

X-RAY OF SPINE (LUMBAR)

MUHAMMAD FAIZ MD SHAKHIH


MMBC 161004
UNIVERSITI TEKNOLOGI MALAYSIA
2

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

CONTENTS
• Introduction
• Materials
• Methods
• Results
• Analysis
• Conclusions
• References
3

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

INTRODUCTION
4

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

Chronic Back Pain


• 84 percent of adults have low back pain at some time in their lives
• Chronic back pain due to fracture, inflammation, herniation, cancer and others.
• Diagnostic technique:
▫ Non imaging (routine) –
▫ History taking (fall/heavy work/medications)
▫ Physical examination (neurological examination, palpation of spine, inspection of posture),
▫ Lab analysis (source of infection, cancer biomarkers)

▫ Imaging – Non recommended unless present with red flags symptoms (neurological)
▫ X-ray – Early screening, fracture, mass, bone alignment
▫ CT scan – bone lesion
▫ MRI – soft tissue structure (disc and nerve)
5

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

• Imaging Modalities : Recommended when presented with red flags symptoms


▫ No consensus on gold-standard imaging techniques
▫ Widely available, inexpensive, quick, low radiation compare to CT
▫ X-ray provide information about the alignment, bones structure, cartilage and soft tissue of spine.

• Problem Statement (what do you measure):


▫ Alignment of vertebral bodies (scoliosis, kyphosis, lordosis)
▫ vertebral size
▫ spacing between discs (degenerative disc)
▫ paravertebral soft tissue (widen due to haematoma)
▫ Opacity (osteoporosis)
▫ abnormalities such as fracture (fracture line), mass (metastatic cancer).
6

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

• Objective(s) :
▫ To develop codes that can extract the spine from other structure.
▫ To analyse the anatomical structure of the lumbar spine including it surrounding soft tissue.
▫ To look for any possible abnormalities such as fracture line, mass and others.
7

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

MATERIALS
• Image processing techniques

▫ Image acquisition : X-ray of Lumbar in AP projection


▫ Image pre-processing : Contrast enhancement, Homomorphic
filtering
▫ Image segmentation : Image cropping
▫ Feature extraction : Thresholding, K-mean clustering, Labelling
8

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

Flow chart & SOUCE CODE


9

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

THRESHOLDING K-MEAN HOMOMORPHIC

Image acquisition Image acquisition Image acquisition

Cropping Contrast enhancement Contrast enhancement

Contrast enhancement K-mean selection Homomorphic filter

Pixel intensity looping Label selection Cropping

Holes filling Cropping Labelling


10

Thresholding Technique MMBC 1093 Biomedical Imaging


and Image Processing
1/9/2019

Insertion of image Cropping Image Pixel intensity looping Holes filling


enhancement
X= imread('lspine2_2.jpg'); crop=X; y=adapthisteq (crop); H1=150; M=imfill(p);
X= rgb2gray (X) crop(1:65,:)=0; figure, colormap (gray), L1=0; figure, colormap (gray),
crop(731:1251,:)=0; imagesc (y) imtool(y) imagesc (M);
crop(:,1:380)=0; y=adapthisteq (y); for i3=1:10
crop(:,578:1024)=0 figure, colormap (gray), p=y;
; imagesc (y) for i=1:1251
imshow(crop); for j=1:1024
if y(i,j)>= L1 && y(i,j)<= H1
p(i,j)=0;
else
p(i,j)=255;
end
end
end
figure, colormap (gray), imshow
(p,[])
H1=H1+5;
end
11

K-mean Clustering Technique MMBC 1093 Biomedical Imaging


and Image Processing
1/9/2019

Image
Image insertion K-mean selection Label selection Cropping
enhancement
I=imread('lspine2_2.jpg'); %convert to grayscale %k-mean selection %Identify pixel range crop=convert;
I= rgb2gray (I); K=30; H1=0.86; crop(1:65,:)=0;
Ikm2=Km2(rgbImage,K); L1=0.82; crop(731:1251,:)=0;
%apply enchncement imagesc(Ikm2); crop(:,1:380)=0;
P= adapthisteq (I); %label the selection crop(:,578:1024)=0;
rgbImage = cat(3, P,P,P); %convert back to rgb a=convert; imshow(crop);
convert=rgb2gray(Ikm2); for i=1:1251
imtool(convert); for j=1:1024
if a(i,j)<=H1 &&
a(i,j) >=L1
convert(i,j)=255 ;
else
convert(i,j)=0 ;
end
end
end
figure, colormap (gray),
imagesc (convert)
12

Homomorphic Filtering Technique MMBC 1093 Biomedical Imaging


and Image Processing
1/9/2019

Image
Contrast enhancement & Homomorphic filter Cropping Labelling
insertion
x=imread('lspine2_2.j k=adapthisteq(x); eim=uint8(eim); %label selection
pg'); eim=rgb2gray(eim); H1=190;%background
%homomorphic filter crop=eim; L1=148;
%convert to grayscale dim=k; crop(1:65,:)=0;
x= rgb2gray (x); cim=double(dim); crop(731:1251,:)=0;
[r,c]=size(dim); crop(:,1:380)=0; a=crop;
cim=cim+1; crop(:,578:1024)=0; for i=1:1251
lim=log(cim); imshow(crop); for j=1:1024
fim=fft2(lim); if a(i,j)<=H1
crop(i,j)=0 ;
%gamma threshold else
lowg=.96; crop(i,j)=255 ;
highg=1.04; end
end
him=homomorph(fim,lowg,highg); end
ifim=ifft2(him); figure, colormap (gray),
eim=exp(ifim); imagesc (crop)
figure;
subplot(2,3,1);imshow(dim);title('Origional image');
subplot(2,3,2);imshow(lim);title('Natural Logarithm');
subplot(2,3,3);imshow(uint8(fim));title('Fourier transform');
subplot(2,3,4);imshow(him);title('Homomorphic filter');
subplot(2,3,5);imshow((ifim));title('Inverse fourier transform');
subplot(2,3,6);imshow(uint8(eim));title('Final result');
13

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

Results and Analysis


14

Thresholding Technique MMBC 1093 Biomedical Imaging


and Image Processing
1/9/2019

Image
Insertion of image Cropping Pixel intensity looping Holes filling
enhancement

• JPG image • Cropping was • Contrast enhancement • Best image was chosen from 1o • Unable to control the
• Frontal view performed first 2x selections of looping images. intensity of the holes
• High brightness before • Provide better image filling.
enhancement for pixel range • Does not really
for better selection improved the
results extraction due to
plenty of small holes.
15

K-mean Clustering Technique MMBC 1093 Biomedical Imaging


and Image Processing
1/9/2019

Image
Image insertion K-mean selection Label selection Cropping
enhancement

• JPG image • Improved contrast • K = 8 vs K=30 • Used 3 pixel range • Provide


• Frontal view once • Provide details selection values better
• High brightness • Enhancement before • Require more pixel range structure
cropping provide identification • Many small
better result for k- holes.
mean selection.
16

Homomorphic Filtering Technique MMBC 1093 Biomedical Imaging


and Image Processing
1/9/2019

Image insertion Image enhancement Cropping Labelling

(a) Contrast enhancement (b) Homomorphic filter


• JPG image • Improved contrast once before applying • Detailed structure
• Frontal view homomorphic filter. • Only few small holes.
• High brightness • Homomorphic filter enhance the intensity • ‘solid’ structure
differences hence make it easier for
thresholding.
• The filter is based on the gamma range value.
17

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

CONCLUSIONS
18

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

Thresholding K-mean clustering Homomorphic filter


Final product

Similarities • Both have almost similar define edges and structure


• Both have missing structure (pedicle)
• Able to extract spine from other overlapping organs
• Soft tissue not visible
Advantages • Fast processing time • Complete and solid structure • Medium processing time

Disadvantages • ‘Porous” • Long processing time • ‘solid’ or complete structure

Suggestions Active Shape Models (ASM) work on shapes, learned


from training images, and then they try to locate the shape
in a test image.

Artificial Neural Networks (ANN) for segmentation (machine


learning).
19

MMBC 1093 Biomedical Imaging 1/9/2019


and Image Processing

REFERENCES
• [1] Ablyazov, O. (2012). X-ray parameters of lumbar spine. Medical Health and Science Journal, 10(1), 37-40.
• [2] http://tomf.org/gd-resources/downloads/Coury1.pdf
• [3] https://www.uptodate.com/contents/evaluation-of-low-back-pain-in-adults
• [4] https://www.uptodate.com/contents/evaluation-of-low-back-pain-in-adults
• [5] https://dontforgetthebubbles.com/c-spine-x-ray-interpretation/
• [6] https://www.radiologymasterclass.co.uk/tutorials/musculoskeletal/x-ray_trauma_spinal/x-ray_fracture_end
• [7] Stolojescu-CriŞan, C., & Holban, Ş. (2013). A comparison of X-ray image segmentation techniques. Advances in
Electrical and Computer Engineering Engineering, 13(3).
• [8] Nirmala Devi, S., & Kumaravel, N. (2008). Comparison of active contour models for image segmentation in X-
ray coronary angiogram images. Journal of medical engineering & technology, 32(5), 408-418.

Anda mungkin juga menyukai