Tracking Objects of Interest from CT MR Data using

Tracking Objects of Interest from CT/MR Data using Dynamic 3 D Ultrasound http: //ssim. eng. wayne. edu/ssimi/cares/classes/group 7. html Group 6: Michael Shetliffe Mohammad Yaqub Mohammed Alam

Outline Problem Statement Background & Significance Overall Aims Methods n n n Ultrasound MRI Registration Results Problems Solutions Competition Future Work Conclusion

Problem Statement

Problem There is a critical need to update information based on changes occurring during surgery. Changes: n n n Shift Deformation Vascular movement

Overview of Proposal The overall objective of this proposal is to show feasibility and develop a cost-effective and efficient approach to monitor and predict deformation during surgery, allowing accurate, and real-time intra-operative information to be provided reliably to the surgeon. The central hypothesis is that deformation can be followed intra-operatively using ultrasound technology.

Background & Significance Approaches to update the pre-op data: n Intra-operative MR Obtain updated (lower-resolution) MR data Costly, significant setup time, OR compatibility considerations n Finite Element modeling of surrounding tissues Challenges of parameter estimation Parameter variation with physiologic changes Long computation times n Use 3 D Ultrasound Convenient, safe, cheap

Overall Aims Image Acquisition CT US MR Segmentation Registration Evaluate the effectiveness, accuracy, and usefulness of any techniques that were used Testing the method

Image Acquisition – 3 D Ultrasound

What is Ultrasound ? High-energy sound waves (ultrasound) are bounced off internal tissues or organs and make echoes. (2 -13 MHz) Pros: n n Non invasive Real time imaging Cons: n n Cannot image bony structures Poor native resolution Image depends on time to echo (pixel position) and echo strength (pixel intensity)

Ultrasound Images (Breast) Spurious Artifacts Objects of Interest Characteristic US Noise Bench Surface Reflection

3 D Ultrasound – Background 3 D Ultrasound: n n n Tracks 2 D ultrasound probe to build 3 D volume from ultrasound “slices” Can use conventional, portable ultrasound equipment to obtain 3 D volumes Still relies on ultrasonic backscattering from tissue structures intrinsically noisy Image Processing becomes especially important

3 D Ultrasound - System “Stradx” Software 2 D Screen Frames / sec & Probe Motion 2 D Ultrasound Probe 3 D Position Slices in 3 D Space Polaris Tracking System

3 D Ultrasound - Output from Stradx: n n . sx file: Text-based file containing 3 D positions, orientations, sizes, times, etc. of individual 2 D slices. sxi file: Binary file of 8 -bit pixel values What we need: n n 3 D image data in “conventional” format (e. g. dicom, Analyze) that can be read into other systems for processing (Prior to starting this project, it was thought that this was what we would have. )

3 D Ultrasound - Output Examples: 1 -Pass Probe Motion

3 D Ultrasound - Output Examples: 2 -pass had problems with “inter-pass” alignment 2 -Pass Probe Motion

3 D Ultrasound - Post Processing

3 D Ultrasound - Conversion 2 Additional Tools (Provided as part of the Stradx distribution): n Select. SX: (adjust for data too “dense”) n Stack. SX: (create evenly spaced, uniformly aligned slices)

3 D Ultrasound - Conversion Read new. sx file as “Raw” 8 -bit data into a medical imaging tool (xmedcon, osirix, 3 d-slicer) Region of Interest moves within resliced image between slices.

3 D Ultrasound - Conversion At this point, we can treat our US data as a standard set of images, and export to other convenient formats Slice separation distance determined from. sx file position data.

3 D Ultrasound – Post-Processing Basic Objectives: n n Reduce background noise Segment (or at least highlight) object(s) of interest Approach: n n Stradx has some inbuilt segmentation/registration capabilities – not used Used MATLAB to investigate feasible techniques that could later be integrated into a stand-alone system.

3 D Ultrasound – Post-Processing Noise Reduction Techniques: Original Image Median Filter Gaussian Blur

3 D Ultrasound – Summary In this portion of work we have: n n n Acquired updated ultrasound image data that better meets our requirements for working with objects of interest. Converted the image data to formats that can be easily used in a wide variety of medical imaging systems. Processed the ultrasound image data to enable higher quality results from subsequent registration with other modalities.

Image Acquisition MRI / CT

What is MRI? MRI stands for Magnetic Resonance Imaging. The MR images used to image internal structures of the body, particularly the soft tissues. We used GE 4 tesla MRI machine to get some MR images for the phantom we have. MR brain image

GE MRI machine

What is CT? CT stands for Computed Tomography. CT image is a specialized form of x-ray imaging. It shows bony structures. We used CT images in our project. It did not give good results. CT brain image

MRI/CT Images (Breast) CT Images No internal information MR T-1 Images MR T-2 Images No big difference because the phantom is MR incompatible

Image Segmentation MRI

MRI Segmentation No quantitative phantom information. Introduce artificial shift of objects of interest. n We did the movement using a manual linear interpolation method. Segmentation n n Object matching Thresholding

Example Original MRI slice Manually shifted object of interest

Example (cont. ) Original segmented object of interest inside an MR image Automatically segmented object of interest inside the shifted MR image

Image Registration

Registration Registering different modalities: n n n Original MR to shifted MR Original MR to CT Original MR to Ultrasound Manual landmark. An automated registration - mutual information. Need more human interaction.

Example Original fixed modality Moving modality (The Shifted object of interest)

Example (cont. ) Original object of interest shifted object of interest A slice that contains the registered data Showing both the original & the shifted points

Example (cont. ) MR image with three objects Ultrasound image with three objects Need for deformable registration method

Problems, Proposed Solutions and Future Work

Problems Phantom n n Echo-sensitive material Water Content MRI Compatible Atlas Ultrasound n n Quality of Images System Calibration Registration

Proposed Solutions - Phantoms Multi modality Phantoms - $ 2000. 00

Proposed Solutions - Ultrasound Better Ultrasound Transducer

Proposed Solutions - Registration Existing Research

Competitors Ultrasound Probes - Philips, Siemens, GE Registration Techniques – Related research available Overall Intraoperative Tracking System using Ultrasound – GE Overall Intraoperative Tracking System using Ultrasound +preop MRI - ? ?

Future Work Research areas n n n Better post processing of US images Newer registration techniques Faster and effective calibration methods Bring all individual modules to work as a single system Testing and evaluation techniques

Conclusion A very feasible and highly applicable research area. n n n Cost effective when compared to I-MRI Relatively “real time” data Accuracy from fusion with pre-op MR Improved surgical outcome Works in current clinical settings Adaptable to other surgical procedures involving: Brain Breast Prostate Others …

Thank You! http: //ssim. eng. wayne. edu/ssimi/cares/classes/group 7. ht ml