1. Brain Image & Lesions Registration and 3D Reconstruction in DICOM MRI Images
C.P. Loizou1, C. Papacharalambous1, G. Samaras1, E. Kyriakou2, T. Kasparis1, M. Patziaris3, E. Eracleous3, C.S. Pattichis4
1Cyprus University of Technology, Department of Electrical, Computer Engineering and Informatics
2Frederick University, Department of Computer Science, Limassol, Cyprus
3Ayios Therissos Medical Diagnostic Centre, Nicosia, Cyprus
4University of Cyprus, Department of Computer Science

2. Outline
Introduction-Multiple sclerosis (MS), 3D Registration & Reconstruction
Objective
Materials & Methods
Study group, Phantom and MRI Image Acquisition
MRI image preprocessing and filtering
Lesion Segmentation
Registration & Reconstruction
3D Volume Estimation
Evaluation Metrics & Statistical Analysis
Results
Discussion, Concluding Remarks, Future Directions

3. 1 Introduction-MS, 3D Registration, Reconstruction
MS: Refers to scars particularly in the white matter of the brain with or without neurological symptoms. [1]-[4]
3D Imaging systems follow up the development of the disease [1]-[4]
[1] C.P. Loizou et al., Quantitative texture analysis…, J Neuroarad, 2015; [2] N. Chumchob, et al., ‘A robust affine image..”, Int. J. Numr. Anal. & Model., 2009; [3] C. Kumar, et al., “3D Reconstruction of brain tumor from 2D MRI’s … ”, Int. J. Advanced Research Electron. & Commun. Engin., 2014; [4] M. Arakeri, et al., “An effective and efficient approach to 3D… ”, Int. J. Signal Proces. & Image Proces., 2013.

4. 2 Objective-Motivation
Propose/evaluate a brain MRI registration and 3D reconstruction system for the 3D reconstruction of DICOM brain images and lesions from MS subjects. Validation on calibrated 3D MRI models and real DICOM MRI brain images and MS lesions.
Motivation:
Prevention is better than cure
Individuals at risk can be identified

5. Flow diagram analysis of the proposed registration and 3D
reconstruction method in DICOM brain MRI images.
[1] C.P. Loizou et al., Quantitative texture analysis…, J Neuroarad, 2015

6. 3.1 Materials & Methods-Study group & Dicom MRI Image Acquisition
ACR MRI Phantom (17x17x10 mm rectangle grid, cylinders and other points of interest)
T2-weigted MRI DICOM using the Phillips scanner
Acquisitions at two different time points (T0 and T1)
American College of Radiology (ACR) MRI Phantom [1]
56th slice of the ACR MRI phantom
Phillips Achieva MRI scanner 3.0 T
[1] American College of Radiology. Phantom Test Guidance, 2005,

7. 3.2 Materials & Methods-MRI image preprocessing, filtering, lesion segmentation
Image histogram equalisation prior reconstruction [2]
Wiener filtering [3]
Manual segmentation of lesions by an MS neurologist and confirmed by another radiologist
[2] C. P. Loizou, S. Petroudi, I. Seimenis, M. Pantziaris, C.S. Pattichis, “Quantitative texture analysis of brain white matter lesions derived from T2-weighted MR images in MS patients with clinically isolated syndrome,” J. Neuroradiol., vol. 42, no. 2, pp , [3] M. Martin-Fernandez, et al., “Sequential anisotropic Wiener filtering applied to 3D MRI data,” Magn. Res. Imag., vol. 25, no. 2, pp , 2007.

8. 3.3 Materials & Methods-MRI Registration, 3D Reconstruction and volume estimation
Time points T0 and T1 were registered using a non-rigid image registration method [8]
Generate a 3D volume using iso-surface rendering [9], [10]
MS lesions were also registered and reconstructed similarly
Volume estimation [9]
[8] J. P. Thirion, “Image matching as a diffusion process: … ,” Medic. Image Anal., vol. 2, no. 3, pp. 243–260, 1998.
[9] R. M. Sherekar, A. Pawar, “A MATLAB image processing approach …..”, Americ. J. Mechan. Engin. and Autom., vol. 1, no. 5, pp , [10] C. Koniaris, et al., “Survey of Texture mapping techniques ….”, J. Computer Graphics Techniques, vol. 3, no. 2, pp , 2014.

9. 3.4 Materials & Methods-Evaluation Metrics
Following evaluation metrics were used to assess the proposed registration and 3D reconstruction methods [1]:
True positive rate (TPR),
True-negative rate (TNR),
False positive rate (FPR),
False negative rate (FNR),
Accuracy (ACC%),
Mean square error (MSE),
Correlation coefficient (ρ),
CXentroid (C),
Perimeter (P) and
volume (V).
[1] C.P. Loizou et al., Quantitative texture analysis…, J Neuroarad, 2015

10. 4.1 Results – Phantom and MRI Registration

11. 4.2 Results – Statistical Evaluation 1

12. 4.3 Results – Statistical Evaluation 2

13. 5. Discussion
No other studies found in the literature for 3D MRI brain registration & reconstruction on DICOM images
3D Slices and 3D lesions accurately reconstructed and inserted to the brain volume
Integrated software and may be used to detect, extract and follow-up MS lesion in MS subjects in longitudinal studies,
Analyzing the evolution of MS and their impact on surrounding structures.
Used for treatment planning, therapeutically monitoring, surgery and pathological brain modelling.

14. 5.1 Conclusion and future directions
The evaluation showed that the proposed method may also be reliably used in the registration and 3D reconstruction of brain MRI images. However, further work in a larger number of images, as well multiple observers’ evaluation is needed for further validating the proposed method for the clinical practice.

15. Thank You