Infrastructure for IGRT [bits, bytes, and but(t)s] Bruce Curran University of Michigan Ann Arbor, MI Disclosures • My department has research affiliations with – Varian Medical Systems – Calypso What is “Infrastructure”? … most generally, is the set of interconnected structural elements that provide the framework for supporting the entire structure. It usually applies only to structures that are artificial. The term is used differently in a variety of fields; perhaps the single most well-known usage is in economics, where it refers to physical infrastructure such as buildings and roads. From http://en.wikipedia.org/wiki/Infrastructure Infrastructure The notion that a structure has an in-ternal fra-mework is popular especially in business organizations where a dependency on interconnected information technology systems has become as prevalent as a city's dependency on interconnected conveyance systems for power, people and things. From http://en.wikipedia.org/wiki/Infrastructure What’s in an “Infrastructure”? • Computer Systems – Storage – Visualization – Analysis Tools • • • • Networking / Connectivity People Resources Machine Resources Training State of IGRT • Weekly Portal / Orthogonal Imaging – Bony Anatomy Alignment / Review – Marker Alignment / Review • Daily Portal / Orthogonal Imaging • Weekly Cone-Beam CT – Soft Tissue Alignment • Daily Cone-Beam CT • Other Imaging / Positioning Modalities – – – – US Infrared Video RF, Other positioning signals Sample Intra-Fraction Prostate Movement AP IS LR Position (cm) 0.0 -0.2 -0.4 0 60 120 180 240 0 60 120 180 240 300 360 420 480 540 600 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 300 360 420 480 540 600 Time (s) Courtesy of James Balter Impact of prostate positioning strategies (Litzenberg RSNA 2005) PTV Margin (mm) 12 10 IS AP LR 8 6 4 2 0 Skin Marks Initial skin-based setup Initial Setup to prostate Pre-Beam Correction Intrabeam correction Our Model Department • Treats 40 patients / day / linac – (10,000 treatments / year) • 3 linear accelerators • One CT Simulator • We’ll ignore for the moment – Brachytherapy – Multimodality Imaging Studies Computer Systems: Storage • Initial Imaging – CT Study (150 images) = 75 MB* • Inhale/exhale/free breathing imaging – 150 Images x 3 = 225 MB • 4DCT – 300 Images x 10 bins = 1,500 MB *Note: Our CT department currently averages 500 slices per CT exam Computer Systems: Storage • So, for our hypothetical Clinic – 70% have a single imaging study – 20% have an inhale/exhale study – 10% have a 4DCT study Leads to 99,000 MB / year / linac or 297 GB / year Computer Systems: Storage • Portal Images (weekly imaging) – ~512 x 512 x 2 bytes = 0.5 Mbytes/image – 4-8 images on first day of treatment – Weekly orthogonal Imaging – And the answer is 2,800 – 3,600 MB / year / linac 8.4 – 10.8 GB / year Computer Systems: Storage • Portal Images (daily imaging) – ~512 x 512 x 2 bytes = 0.5 Mbytes/image – 4-8 images on first day of treatment – Daily orthogonal Imaging – And the answer is 10,800 – 11,600 MB / year / linac 32.4 – 34.8 GB / year – More realistically, 30% are daily imaged 5,215 – 6,000 MB / year / linac 15.63 – 18.0 GB / year Computer Systems: Storage • Cone-Beam CT Imaging (weekly) – ~512 x 512 x 2 bytes = 0.5 Mbytes/image – Image Volume 512 x 512 x ?? = • 55 slices = 27.5 MB • 256 slices = 128 MB – Image 20% of patients = 11,000 - 51,200 MB / linac / year 33.0 - 153.6 GB / year Computer Systems: Storage • Cone-Beam CT Imaging (daily) – ~512 x 512 x 2 bytes = 0.5 Mbytes/image – Image Volume 512 x 512 x ?? = • 55 slices = 27.5 MB • 256 slices = 128 MB – Image 10% of patients = 27,500 - 128,000 MB / linac / year 82.5 – 384.0 GB / year Computer Systems: Storage • Summary: – 50% (weekly orthogs): – 30% (daily orthogs): – 10% (weekly CBCT): – 10% (daily CBCT): • Total Verification (per linac): • Imaging (per linac): • GRAND TOTAL (clinic): 1.60 GB 3.36 GB 15.00 GB 75.00 GB 94.0 GB 99.0 GB 579.0 GB – x 3 for backup / RAID: 1.737 TB/year But ….. • Where do we store this information? – CT Simulator Archive – Planning System – Treatment Management System – ... So you may end up with 3 or 4 times the original information depending upon your practice, and you may not be able to archive it per year (some systems will not allow archiving of partial data), and we haven’t talked about saving the CBCT projections for research or other investigations (estimates from Sonke were about 10x the CBCT size!) And … • Currently Radiology at UM is averaging a 50%/year growth in imaging information (total exams are relatively flat) • As we learn more about how to use IGRT to reduce PTV margins, we will most certainly increase the number of advanced (e.g. 4D or CBCT) scans we utilize in planning and treatment But … • We will likely have some deformation maps for our inhale/exhale and 4DCT data In the worst case, these could be {x,y,z} vectors, one per pixel, so LEAVE SOME EXTRA ROOM!!!!!!!!! Visualization • What tools will we need to efficiently align patients while under treatment – Match and Adjust – 2D – 2D matching – 3D – 3D matching – Registration / Fusion • Proper CRT/LCD and lighting at linac console Match and Adjust Match and Adjust • Works with markers, bony anatomy • Can require experienced RTT for accurate use • Filtering may be required to highlight anatomy (diaphragm) • Needs someone to indicate the appropriate anatomy for matching 2D-2D Match 2D – 2D Match • Using kV imaging, gives higher contrast images – Allows true lateral imaging for markers in pelvis • Reduces RTT skill necessary for accurate match • Still limited to high contrast objects 3D – 3D Match 3D – 3D Match • Allows soft tissue matching, e.g. GTV • Can generally reduce errors to those of ITV • Tools, time often not yet quite there – Additional time / training Networking / Connectivity • How long are you willing to wait? – Server throughput – Network speed – Workstation speed • Some Recommendations – 1 GB network connectivity at server • Additional network cards in a busy department – Fast Disk Systems • 10,000 RPM SCSI or SATA – Pentium 4 or equivalent • As much speed, particularly on I/O bus, as you can afford Networking / Connectivity • Dataset Transfer Time – 100 MB Image Dataset = ~ 1 Gbit of dat – Using an 100 Mbit ethernet • • • • 10 Seconds? Overhead of about 20% Typical efficiency of about 30% Result is about 30 – 40 s BUT …. Networking / Connectivity • Network efficiency drops rapidly as network traffic increases – At 60% utilization, efficiency may be as low as 10% (collisions due to network algorithms) – Delay becomes highly unpredictable – Thus the 30 – 40 s may actually end up as 3 4 minutes or more • Routers / firewalls add additional latency People Resources • Off-line Analysis – Need for additional RTTs/dosimetrists • On-line Analysis – Who approves the move? When? (Reimbursement is muddying these waters!) • MD Resources – More images / anatomy to review • Physics Resources – QA / Analysis / Protocol Development • IT Support Machine Resources • TIME!!!!! Training • Team Approach – Develop procedures and modify as needed! • Typically train 1-4 RTTs for initial implementation – QA runs are useful for procedure familiarity • Documentation / References Linac 4DTC 1 Mode Up OBI Setup Field OBI WS 2 Click 2D/2D Match 5 Analyze 3 Send Field and Choose Technique Press Motion Enable Foot Switch to Acquire Manual Matching Review new couch Positions 4 Send Next Field and Choose Technique Press Motion Enable verify 6 Save couch position for current session or all sessions. Send in first field Foot Switch to Acquire Send couch position and treat or verify with imaging treat Press Motion Enable Beam On Courtesy of Scott Hadley Acknowledgement Control Systems • Paul Keall said it all!!!!! Discussion