Elliot K. Fishman M.D., FACR, Professor of Radiology and Oncology, Johns Hopkins University School of Medicine, Baltimore, Maryland
The medical headlines on CNN were most impressive on that early fall morning in September 2003. The announcer described a new non-invasive imaging technique developed at the same English company that in the past had brought us such musical superstars as Brittany Spears and Eminem (note: I had to update singers for relevance as the Beatles are so yesterday) just introduced a scanning device that goes by the acronym of a CT scanner, short for Computed Tomography. Using the latest hardware and computer technology the scanner is in constant motion and rotates around the patient every 500 milliseconds. The scanner allows 16 individual scans or slices to be obtained per scanner rotation or 32 scans per second. The system can acquire isotropic data where the x, y, and z resolution is equal. Typical scanning parameters used are. 75 mm slice thickness, .5 mm interscan spacing and post processing of data which allows reconstruction of 6 slices per second. The speed of acquisition means that a pancreas, liver or kidney can be evaluated with isotropic datasets in under 10 seconds. Depending on the study between 500 and 1200 individual slices are obtained for review with some studies such as vascular run-off studies generating 1500-2000 slices. The manufacturer noted that future generations of the scanner will have 32-256 detectors before the next wave of technology goes to flat panel detectors.
The story continues and the announcer then goes to a leading medical institution and shows the images to a startled and awed home viewing public. Now for my question. Does anyone actually believe that the images demonstrated would be axial CT scans or are they more likely to be images from a volumetric 3D display. Does anyone believe that an axial display would even be a consideration if this new system were indeed introduced today? I doubt it. Yet, most of us still look at our shiny new 16 slice scanner and review axial images. However, the times are changing.
As we approach the 30th anniversary of the introduction of Computed Tomography (CT) into clinical practice, CT technology continues to evolve and is thriving. In the early years of CT, technical advancements were typically measured by decreasing scan times and increasing speed of data reconstruction. Individual study times decreased from 90 minutes in the early 1980’s to less than 20 minutes by the end of that decade. Although this increased speed definitely represented progress, the first truly revolutionary concept change in CT occurred in the late 1980's with the introduction of helical or spiral CT. This technology along with significant advancements in image post-processing software revitalized CT and allowed the development of new CT applications such as CT angiography and virtual imaging (1-3). The era of true volume imaging had begun in earnest.
Next, the transition from single detector to multidetector row CT represented another pivotal event in the evolution of CT imaging. Initially 4 then 8-detector scanners were developed. But, it was the introduction of 16 detector scanners, which has really revolutionized CT scanning and requires changes in our core operative structure. 16-slice MDCT represents what Andrew Grove, Ph.D. (co-founder and former CEO of Intel Inc.) called a strategic inflection point. It is defined as a change, which is not a 5-10% improvement over previous capabilities, but one that strategically changes the whole landscape. 16 slice CT is not just simply 4 times faster than 4 slice scanning. It is far more than an incremental improvement. 16 slice MDCT, with true isotropic datasets, promises to further advance volume imaging and all its resultant advantages. Volume imaging will no longer be considered as a supplement to traditional axial imaging in select cases, but instead, may now be utilized for primary display and analysis. This paradigm shift will require a rethinking of many of the core processes in CT ranging from image transfer and data storage to redefining the role of CT across a wide spectrum of clinical applications. For example, the coronary arteries or peripheral runoff studies were once considered possible but impractical with 4-slice technology. These applications are now a reality (4-7). Workstations like the 3DVirtuoso, which worked so well with single slice or even 4-slice MDCT can not handle these new volume datasets and provide real-time interactivity. Workstations are no longer a simple accessory to the CT scanner but the central core of processing and display.
CT continues to prosper in the medical imaging marketplace today with a growth rate of 10-15% in yearly CT volume. The new scanners have increased throughput capabilities to at least 3-4 patients an hour with the limitations being the time it takes to place the patient in the scanner and prepare them for a study. The problems of tube heating or slow data processing are for the most part a thing of the past. Many of these cases are more complex and require multiphase acquisitions. These increased capabilities coupled with increased experience can result in 3000-5000 images (or more) being generated per hour. A specific example is the 4D cardiac studies that require reconstruction at 9 intervals (10%-90% of the cardiac cycle) which represents 3000 images or so in a single exam. This total volume of data is in the 30-50 gigabyte range per day and the challenges that this presents have been the subject of many articles in imaging journals and in radiology web chat rooms.
The solution to the problem of large volume visualization with 4- and now 16-slice MDCT is not simply a modification of standard techniques such as faster scrolling with a computer mouse or roller ball or novel filming protocols (i.e. film every 5th or 10th image). Rather, it focuses on the data volume itself with the understanding that volume acquisition requires true volume visualization. The introduction of programs such as InSpace on the Siemens Sensation 16 scanners and on the Leonardo workstation addresses this paradigm shift from axial images to volume images. The study volumes can be presented in a range of formats including volume rendering, maximum intensity projection, minimum intensity projection and multiplanar visualization (coronal, sagittal, axial) .
InSpace allows primary analysis of the data volume in real time which is a necessity given that it is no longer practical to review CT slices on film or even using a film surrogate, the PAC's workstation or review station. For instance, with a 4-slice MDCT a dual phase dataset using 1.25 mm thick sections of the liver or pancreas was composed of 400-500 sections. At that point, it was still possible, although cumbersome to scroll through these images on standard workstations. Yet, in fact, most centers reviewed images at 5-mm thickness and 5 mm intervals and at best looked at the 1-mm sections as part of the 3D examination. Regardless, this strategy of image review is not possible with 16- slice MDCT scanners. With the use of the Sensation 16 scanner we are performing detailed CT angiographic work with .75 mm thick sections reconstructed at .5 mm intervals. The optimal reconstruction interval overlap is .25 mm, which provides a dataset, which is ideal for 3D reconstruction and multiplanar imaging. If we scan an area like the pancreas we are typically doing arterial phase images followed by venous phase imaging. This results in 400-500 individual slices per acquisition. If you view this dataset on a workstation or film it is likely that you only reconstructed data at 3 or 5 mm intervals which means you are looking at between 10 and 16% of the available information. Although I would agree that in many cases the additional slices do not change the specific diagnosis, they may change the staging of disease especially in regards to vascular invasion. The concept that must be understood is the ability to use a 16-slice scanner in the optimum mode requires a new workflow strategy. This shift to volume display not only affects the radiologist at time of interpretation of the study but also the referring physician at the time when they review the study.
The impact of the changes due to 16 slice MDCT has not taken the referring physician needs into the equation. Imagine their frustration looking at x-ray jackets with 20-100 CT sheets of film trying to find the key image. In many ways for our colleagues more is simply less. I would conclude that unless handled correctly more information (i.e. more slices per study) might provide less diagnostic information due to a breakdown in information transfer. This is a critical problem and one our colleagues are becoming frustrated with and one that can be addressed with new visualization strategies.
This paradigm shift to volume visualization has some challenges to overcome. Although a solution for volume image viewing has been addressed with InSpace, the ability to move images through the radiology department as well as the hospital or health care enterprise remains a challenge. Similarly, the amount of data generated provides perplexing problems for data storage and retrieval. Studies that range in size from 600 megabytes to 1 gigabyte are truly a challenge that must be addressed. New solutions must be identified and implemented. The size of the datasets per case make it no longer possible to use 5.2 gigabyte Sony MO storage devices as these hold but a handful of cases (usually 3-8 depending on use of image compression). This process as currently used is the worst of all worlds; it is expensive, time consuming, and inefficient. Yet, most 16 slice scanners still ship with 5.2 G MO storage devices. A new workflow design with massive storage is necessary either as part of the hospital or clinic master plan (institutional PAC's system) or as a freestanding enterprise solution within the CT environment.
Even with a central archive not all issues are resolved. We require rapid data retrieval which may be limited by network transfer times with limited bandwidth which can paralyze the entire operation. Although this is commonly felt to be a problem for the Radiology department, it is, in truth, a problem for the scanner vendors. Simply supplying "data acquisition devices" like a CT scanner is no longer enough as it alone is neither a solution or a system. Scanner manufacturers need to address workflow and process design. Siemens Medical Systems seems to have recognized the future in its name change to Siemens Medical Solutions, but now comes the hard part. Name changes are easy but providing those solutions that live up to your name is the challenge. Similarly, GE Medical is focusing their operations as providing system solutions rather than hardware solutions. Time will tell if they are successful. We can only hope they are. Recently GE Medical has teamed up with EMC, the leader in enterprise storage solutions. We look forward seeing any novel solutions this may bring to our clinical environment.
Another issue that
relates to workstations like the Leonardo is the amount of system storage available.
These systems typically have hard drives in the 60 gigabytes range (55 G to
be exact). This is insufficient in an environment where 20-30 cases per day
are in need of analysis whether it be CT angiography, 3D imaging, virtual colonoscopy
or whole body screening (8-10). I recently purchased from Apple Computer
a 400-gigabyte hard drive for 700 dollars. This should be the bare minimum
on a workstation with 1-3 terabytes probably a better number. More sophisticated
storage strategies should also be made available for sites like ours where we
typically keep months or years of CT angiographic and 3D online. I would
guess a system with 10-40 terabytes storage is not undoable and would be cost-effective.
This need for massive local storage should come as no surprise to anyone. The
problem is that most of the major vendors are designed so that the people who
do system design do not have an understanding of what the enduser needs or wants.
I have been told several times by major vendors that the problem of limited
storage is not the manufacturer’s lack of understanding of the clinical environment
but mine. Enough said.
Another problem with some current workstations is that they have not taken the clinical environment into their planning process for their database design. The databases are not flexible enough to allow user selected partitions such that case studies could be divided by topic (i.e. pancreas, liver) as well as being limited in how many absolute cases can be within the database. Although some might be tempted to suggest that a workstation was never meant to store cases this is again not a real world analysis or solution. As a user I need a minimum of 4-6 weeks of 3D cases, and ideally at least 3 months on line. Clinicians will commonly want to review studies done in the past 4-6 weeks or longer (especially for renal donors) and they must be available. The limitations currently provided are real and not theoretical. As of March 2003, we had three 16-slice MDCT’s feeding a workstation in our 3D imaging lab. The volume of data results in a purging of the database as least twice a week. This is both time consuming and interfering with delivery of services. The concept of a strategic inflection point could have predicted problems of this scale and magnitude.
Another issue common in many institutions is the need to minimize the time from data acquisition to when the radiologist has the ability to read the case. This problem may vary and will be different depending on the scanner and its configuration. If a Siemens Wizard (satellite console) or Leonardo workstation is used as the second console and is dedicated to physician review then the shared database with the main scanner provides easy access to current scan data. This is very efficient for throughput and consultation with the referring physician. However, if it is not available, other solutions including a PAC’s workstation can be used but there will be no shared database. Sufficient network bandwidth is needed to prevent any bottlenecks in this scenario as the information is sent across the network. Radiology, like many other businesses, is becoming ever more dependent on the quality and reliability of our network backbone. It is critical to our ability to efficiently practice radiology in the 21st century.
An area that is commonly overlooked with 16-slice MDCT is the impact these changes in technology are having on our referring physicians. The referring physician is not interested in stacks of images on film or on a workstation. Clinicians demand rapid access to critical information and images displayed in a user-friendly environment. 100's or 1000's of images per patient on film or on a computer screen or computer disc are not acceptable. In contrast, an interactive volume display may be the answer. The combination of a limited number of selected volume visualizations coupled with the capability of real time 3D rendering should prove ideal for a wide range of applications. If our personal experience is any guide, the acceptance of volume displays is immediate, and complete. The physicians come down to the scanner to view the 3D volume displays and several of them have even taken the time to learn how to use the software.
The concept of InSpace or a similar system when available on multiple workstations or PAC's systems across the enterprise moves us one step closer to the true "virtual" radiology department. That is, it matters little where the information is acquired, as it is available for a primary read or review anywhere throughout the enterprise. This enterprise-wide solution is critical to the radiologist and the referring physicians who would also have access to the volume datasets. The quality of a volume dataset is of little value unless the information can be used.
Our experience is that nearly all referring physician but especially surgeons have an increased need for access to our data files. The details provided by 16-slice MDCT are being used for detailed surgical mapping and planning. The need for more interactive tools in the clinic and the operating room are two of the common requests that I am currently seeing. Companies like TeraRecon (San Mateo, CA) have addressed this issue with a thin client environment solution that decreases the cost per seat on the network by 50-75%. It will be interesting to see how well this model catches on. Some radiologists are concerned that sharing data may decrease the radiologist’s control of the information. Although this is in part correct, I believe that if radiology leads this new work paradigm we will remain the central player even in a distributed information environment.
Documentation of the results of 3D imaging or CT angiography can be done in several ways, any which may or may not work in your clinical environment. Images can be filmed on a laser camera like routine CT scans. This can be useful especially if you are still in a film-based environment. What is perhaps better in our experience is to film images directly to photographic film, which is then given directly to the physician or sent to him/her. In the past, we have done this using a Kodak 8650 dye sublimation printer, which makes good quality images at under $3.00 per page. We typically might give the referring physician anywhere between 6 and 14 images or 3 to 7 sheets of film. Recently, a new lower cost camera from Olympus has been introduced in the consumer market but has proven ideal for our lab. The systems cost around $400 and produces images of a better quality than the Kodak 8650, which might cost 20-50x more. Individual film cost is only $1.60 per sheet. Currently it is our standard of communication and well accepted by our referring physicians. Other elegant methods of image transfer to the referring physician are via the web as TIFF or PICT files or via CD's or DVD’s. CD's are especially useful when studies that incorporate motion such as virtual colonoscopy are done. CD’s are limited to 700 megabytes and may not hold an entire study without compressing the images. DVD’s can hold around 4.7 gigabytes so would appear to be a better solution in these cases. An analysis of your own environment is critical in determining the optimal delivery system.
Another important aspect of this brave New World of volume visualization is the challenge of training and retraining staff radiologists and radiologic technologists and perhaps ultimately select referring clinicians. Whether this training is on the CT scanner, a workstation, or on a piece of software that runs on both systems, the training process needs revisiting and redesign. The usual method of training is for an imaging specialist to spend 2-5 days at the site of a new scanner or workstation installation. Additional training may be provided at a central location where the training is typically didactic with some hands-on available. This scenario has worked reasonably well for nearly 20 years but is showing its age. A simple mathematical calculation at an institution like Johns Hopkins is that even a week visit for training is grossly inadequate. When you consider our Body CT division have approximately 20 fulltime technologists, 11 physicians who read CT, and 10 fellows and 24 residents who want to learn CT you can see the dilemma. Hands-on training typically helps but a privileged few. The majority of users rely on second-hand training. The same is true with the 3D workstations where much of the training may be combined with the scanner training. Is it any wonder that over 95% of radiologists and technologists surveyed by this author over the past 3 years have been unsatisfied with their training or workstation expertise.
The problem then reqiures new solutions and new teaching paradigm. E-based teaching has proven successful in other fields and will need to be adapted for radiology. The introduction of www.InsideInspace.com in March 2003 attempts to solve some of these problems. Developed as a website dedicated exclusively to using the real-time 3D program InSpace, the site combines technical information, "how to do it" information, and case studies for use by the users or potential users of InSpace. There is a question and answer section for which answer to the most common questions are provided, as well as an "ask the expert" section which puts the user in direct contact with program developers, support staff and radiologists knowledgeable in system performance and clinical applications. The site also provides lectures as well as allows downloadable presets for 3D rendering developed at leading academic and research centers for users to improve their practice. We believe this represents the first attempt to create a true users community within a medical imaging product. It will be interesting to see how successful the site becomes and whether it becomes the prototype for other products and applications. The site is free to the medical community. Whether this paradigm shift catches on and becomes a standard or ‘crashes and burns’ will be of interest to us all. Our personal experience with our own website, www.ctisus.com suggests that radiologists and technologists are ready to learn via the web and that this area will be an exciting one in the next decade (11-13).
CT continues to make amazing technical advancements and is currently enjoying unprecedented popularity based on its ever changing capabilities. These capabilities provide unique opportunities to improve patient care while at the same time pose unique challenges to maximize the capabilities provided by these new technologies. The workflow that has worked so well for nearly 30 years will need to be revisited and redesigned as we move more fully into this world of volume imaging. The opportunity for change is there, we must just embrace it.
1. Horton KM, Fishman EK. The current status of multidetector row CT and three-dimensional imaging of the small bowel. Radiol Clin North Am 2003; Mar 41(2):199-212.
2. Macari M, Bini EJ, Xue X, Milano A, Katz SS, Resnick D, Chandarana H, Krinsky G, Klingenbeck K, Marshall CH, Megibow AJ. Colorectal neoplasms: prospective comparison of thin-section low-dose multi-detector row CT colonography and conventional colonoscopy for detection. Radiology 2002; Aug 224(2):383-92.
3. Lawler LP, Corl FM, Haponik EF, Fishman EK. Multidetector row computed tomography and 3-dimensional volume rendering for adult airway imaging. Curr Probl Diagn Radiol 2002; Jul-Aug 31(4):115-33.
4. Klingenbeck-Regn K, Flohr T, Ohnesorge B, Regn J, Schaller S. Strategies for cardiac CT imaging. Int J Cardiovasc Imaging 2002; Apr 18(2):143-5.
5. Flohr T, Ohnesorge B. Heart rate adaptive optimization of spatial and temporal resolution for electrocardiogram-gated multislice spiral CT of the heart. J Comput Assist Tomogr 2001;Nov-Dec 25(6):907-2.
6. Funabashi N, Kobayashi Y, Perlroth M, Rubin GD. Coronary artery: quantitative evaluation of normal diameter determined with electron-beam CT compared with cine coronary angiography initial experience.Radiology 2003; Jan 226(1):263-71.
7. Rubin GD, Schmidt AJ, Logan LJ, Sofilos MC . Multi-detector row CT angiography of lower extremity arterial inflow and runoff: initial experience. Radiology 2001; Oct 221(1):146-58.
8. Lawler LP, Pannu HK, Corl FM, Fishman EK. Multidetector row computed tomography with volume rendering-an aid. Curr Probl Diagn Radiol 2002; Nov-Dec 31(6):230-43.
9. Lawler LP, Corl FM, Fishman EK. Multi-detector row and volume-rendered CT of the normal and accessory flow pathways of the thoracic systemic and pulmonary veins. RadioGraphics 2002; Oct 22 Spec No: S45-60.
10. Fishman EK, Horton KM. Imaging pancreatic cancer: the role of multidetector CT with three-dimensional CT angiography. Pancreatology 2001; 1(6):610-24.
11. Scatarige JC, Garland MR, Corl FM, O'Keefe CF, Fishman EK. Visitors and content preferences on an educational web site dedicated to clinical body computed tomography: results of a 2001 audit and online survey.Invest Radiol 2002 Feb37(2):53-9.
12. Corl FM, Kuszyk BS, Garland MR, Fishman EK. 3-D volume rendering as an anatomical reference for medical illustration.J Biocommun 1999; 26(2):2-7.
13. Horton KM, Garland MR, Fishman EK. The Internet as a potential source of information about radiological procedures for patients.J Digit Imaging 2000; Feb 13(1):46-7.