Jane Lowers

Under the best of circumstances, 3T imaging of the prostate with a body coil can approximate the level of detail and sensitivity available at 1.5T with an endorectal coil. Researchers hope that with a 3T-oriented endorectal coil they will finally be able to take advantage of 3T's higher resolution in a challenging portion of the anatomy.

Three studies presented at the ECR Sunday morning compared 1.5T and 3T imaging, and, overall, the researchers gave the nod to 1.5T, based on coil availability. Dr. Dirk Beyersdorff from Charité Hospital in Berlin imaged 24 patients at both field strengths, comparing an endorectal coil at 1.5T with a body coil at 3T. The 1.5T images were clearer, and peripheral and extracapsular lesions were easier to identify consistently.

"Until a 3T coil is available, 1.5T will remain the gold standard," Beyersdorff said. "The endorectal coil is the most important advancement we are waiting for."

A similar study of 30 patients at the University of Modena, Italy, found that 1.5T imaging could identify six of seven extracapsular tumors, compared with five of seven for 3T. The overall sensitivity and specificity for 1.5T were 85% and 83%, respectively, compared with 71% and 83% for 3T. 

Commercial release of a 3T endorectal coil from Medrad is expected later this year, said session moderator Dr. Hedvig Hricak, chair of radiology at Memorial Sloan-Kettering Cancer Center. At the University of Nijmegen, Netherlands, which has worked with an experimental version of the coil, sensitivity for prostate masses increased from 68% to 87% for experienced readers, according to lead researcher Dr. Jurgen Fütterer. 

Despite the presumed clarity of 3T endorectal coil imaging, it won't be the answer for everything, Hricak said. While it may serve an important role in initial staging, body coil imaging will be the preferred technique for planning and monitoring radiation therapy so that the coil does not interfere with mapping the shape and location of the prostate and surrounding tissue.

Articles

Prostate imagers await 3T endorectal coils

Whether it scans in four slices or 16, CT is the method of choice for detecting pulmonary emboli, researchers said at the RSNA meeting. Even four-slice technology boasts a negative predictive value greater than 99%, better than ventilation/perfusion (V/Q) studies or conventional pulmonary angiography. 

Differences among four-, eight-, and 16-slice scanners appear marginal so far, but few data are available to solidify any claims of superiority for 16, except in speed. Researchers at Stanford University clocked lung scan times at 26.6 seconds for four-slice, 9.25 seconds for eight-slice, and 5.45 seconds for 16.

Dr. Alessandro Napoli and colleagues at Stanford reviewed CT studies for 1240 consecutive patients referred for suspected pulmonary emboli. Overall, 20% of cases were deemed positive, with 41% of emboli found in the segmental arteries, 27% in lobal regions, 17% in subsegmental arteries, and 15% on the main branches. 

While there was no statistically significant difference in detection rates among the three generations of scanners, readers said that eight- and 16-slice studies yielded more and clearer diagnostic data. The 16-slice scanner may eventually prove better at finding subsegmental emboli, but only a small number of cases were included in the study.

If a CT study is negative, it is safe to forgo further examination. Researchers at the University of South Carolina conducted a metastudy of 14 published papers with 3283 patients who had undergone CTA to evaluate suspected pulmonary emboli, had negative findings, and had stopped anticoagulation therapy based on those findings. Fourteen developed fatal pulmonary embolism, 23 demonstrated nonfatal embolism, and 52 showed evidence of deep venous thrombosis during follow-up. 

The studies produced 95% to 100% negative predictive value (NPV), with an average of 99.1%. NPV for mortality was 99.4%. Whether the CTAs were performed on single-slice or multislice CT made no significant difference. By comparison, conventional pulmonary angiography has an NPV of 98.4%, and V/Q studies average 88%.

Whether it scans in four slices or 16, CT is the method of choice for detecting pulmonary emboli, researchers said at the RSNA meeting. Even four-slice technology boasts a negative predictive value greater than 99%, better than ventilation/perfusion (V/Q) studies or conventional pulmonary angiography.

Report from RSNA: CTA nails down role of gold standard for pulmonary embolism

How well a patient can count, remember word lists, or perform other basic tests may not indicate how far Alzheimer's dementia has progressed biologically, researchers said at the RSNA meeting. 

The Mini-Mental Status Exam has been the standard diagnostic tool for Alzheimer's disease for years, relying on a series of simple cognitive tests to calculate a score for degree of mental impairment. In 58 patients who also underwent PET or SPECT imaging, however, the MMSE showed little correlation with imaging evidence of advanced disease. 

Many patients with near-normal MMSE scores showed pronounced bilateral defects on imaging, while others with extremely low MMSE scores showed only moderate PET or SPECT perfusion defects, according to Amitha Rao, a third-year medical student at the University of Kansas, who completed the research under the direction of radiologist Dr. Reginald Dusing.

Patients were excluded from the study if they showed evidence of non-Alzheimer's dementias, vascular defects, or other conditions that could mimic symptoms or invalidate imaging findings. Two readers, board-certified in both radiology and nuclear medicine, interpreted the studies.

The lack of correlation between the two types of tests adds muscle to the theory that adding PET or SPECT may speed the diagnosis of Alzheimer's, Rao said. 

PET has secured limited Medicare reimbursement as an imaging agent in dementia cases, but only to rule out non-Alzheimer's dementias that can not be distinguished clinically from Alzheimer's. 

How well a patient can count, remember word lists, or perform other basic tests may not indicate how far Alzheimer's dementia has progressed biologically, researchers said at the RSNA meeting.

Report from RSNA: PET, SPECT findings challenge office test results for Alzheimer's disease

Breast tomosynthesis may be a work-in-progress in the RSNA exhibit halls, but researchers are already considering ways to incorporate computer-assisted detection algorithms to improve its efficacy.

Researchers from the University of Chicago and the University of Michigan presented papers Tuesday morning that evaluated several techniques for reconstructing suspected tumors from the multiple slices of tomographic acquisition or from the compiled volume data set. Both groups worked with the GE Digital Breast Tomosynthesis prototype developed by Dr. Daniel Kopans at Massachusetts General Hospital.

The Michigan researchers produced an area of 0.87 under the ROC curve with 80% sensitivity and two false positives per case, using a set of synthesized data. Looking only at projection views, the area under the ROC curve dropped to 0.82. Both were based on 26 cases with 23 masses, including 13 malignancies. 

"We achieved fewer false positives with 3D views than with 2D, and the information fusion is more effective," said lead researcher Heang-Ping Chan, Ph.D.

Chicago researchers evaluated only projection views. A standard tomosynthesis scan involves 11 low-dose conebeam projections through the breast at 5-degree angles spread over an arc. The projections intersect with each other at various points and allow the software to confirm areas of suspected masses from more than one angle. For the Chicago trial, any region that was intersected by six or more projections was considered a positive finding. Among 21 cases of confirmed breast masses, the team achieved a sensitivity of 86%, but with six false positives per case.

In addition to projection views and reconstructed volumes, CAD could also be developed to look for suspicious findings across the 50 to 80 reconstructed image slices that could be produced for each set of breast data, the authors said.

Breast tomosynthesis attracts CAD research

A radiologist sitting at a workstation views an abdominal CT study. Priors, matched slice-for-slice, accompany the patient's history and recent lab results. As the speech recognition reporting process begins, software evaluates the text and provides cues to speed and direct the rest of the report. The radiologist adds notes to the images that best display the pathology. Based on the radiologist's conclusion, the software offers links to similar cases and medical references. 

The next case, a CT colon study, is queued up and presented in the radiologist's preferred mode, coronal reconstructions. A CAD algorithm has already run and awaits review. 

On the receiving end, the ordering physician reading the report on a modified version of the radiologist's workstation can pull up the key images, relevant notes, and links, as well as the patient's history and recent diagnostics. With a click, a patient-friendly version of the report prints out for the next office consultation. 

This scenario is an improbable one today, but some radiology innovators suggest it's the way things could be done some day-and perhaps the way they should be done if radiology wants to retain its place at the clinical table. For all the training and expertise that radiologists accumulate, the radiology report is for the most part their only interface with the rest of the medical world. Their ability to lend diagnostic clarity boils down to a few hundred words that may reach the referring clinician a day or two after the initial request for imaging is made. 

"The radiology report has always served partially as a PR vehicle, a chance to prove your worth to the rest of the medical enterprise," said Nancy Knight, Ph.D., a radiology historian. 

In radiology's early days, the field risked becoming a hospital service rather than a medical specialty, and reports and face-to-face consultations were the most effective means of convincing clinical peers that radiology required specialization and could provide expert service. 

"For most of the 20th century, referring physicians had to go to the radiology suite to see images, meet the radiologist, and have a conversation about the findings," Knight said. "With film going digital, some observers worry there's a danger of losing that contact, that a second 'crisis of identity' may be on the way." 

A purely digital interaction isn't the only risk. Clinical specialties that are investing in their own imaging equipment may lose sight of radiology's expertise, and local clinical customers may eventually weigh whether to send studies to competitive radiology services in other states or countries. The radiology report, which in most respects contains the same information in the same format as it did a half-century ago, may be poised for evolution, Knight said. 

Technologically, too, the time may have come. 

"Despite being digital, not a lot has changed in PACS," said Robert Cooke, executive director of marketing for network systems at Fuji. "It's still benchmarked against what is possible with film. A great example of this is the use of the term 'hanging protocols.' We need to think about what's possible in the digital world and with automation." 

In a fully wired healthcare world, patient information would flow back and forth from an electronic medical record through the HIS, RIS, PACS, and workstations. Anyone caring for a patient or performing diagnostic work would have at his or her fingertips a full accounting of the patient's history and current care. Taken a step further, patient data would be standardized, aggregated, and designed to be mined for best practice protocols and more efficient care delivery. Research that would take hundreds of hours of file-diving today could be completed at the touch of a button. The radiology report, the diagnostic starting point for many cases, would be a key player. 

If the goal is to make a radiology report more valuable to its intended audience, it makes sense to first ask members of that audience what, exactly, they want. What they want, according to Dr. Annette Johnson, chief of neuroradiology at Indiana University, is accuracy, clarity, completeness, and timeliness. 

In a survey of IU physicians that netted 160 responses, accuracy came out as the top priority, Johnson said. Two-thirds of respondents ranked timeliness as a top problem, although this radiology department's average is a fairly swift 12 to 13 hours. To the surprise of many radiologists, referring physicians also placed great importance on including pertinent negative findings in reports. 

The survey was part of a larger ongoing project to develop structured reporting techniques and determine whether structured reporting facilitates better quality reports than traditional narrative reporting. IU residents will read neuroscience cases in free narrative, and a panel will evaluate those cases for completeness and accuracy. Three months later, half the residents will read the cases again the same way, while half will use structured reporting. The second sets of reports will be evaluated against each other and against the original round. 

Indiana is a beta site for eDictation, a reporting software used in a number of private practices. Its template system is meant to simplify and streamline the reporting process, but Johnson is trying to take it further, setting up cues that will help residents and radiologists provide reports that meet their clinical colleagues' expectations for thoroughness in a more efficient way. 

For MR in cases of clinically suspected stroke, for example, neuroscience faculty said they wanted to know if a study showed hemorrhage, evidence of infarct, or herniation. 

"In training institutions in academic radiology especially, we might benefit from a checklist of relevant findings, so that once we decide the lesion is in the frontal lobe, then we get a list of findings descriptors. Once we decide which pathology is most likely, we get a list of possible pathologies," Johnson said. "When it's done and you say you think it's ischemia, the checklist should remind you to describe all of the relevant findings-including pertinent negatives-for the benefit of the clinician." 

Johnson has completed such prompts for brain and stroke studies and says the same process could be applied to other subspecialties. 

"One great advantage of structured reporting over voice recognition is that we are able to search for any abnormality or diagnosis easily within these data that are coded when entered," she said. 

The Mayo Clinic took a different approach to putting tools in clinicians' hands with a homegrown image viewing system. All physicians, including radiologists, can add notes or measurements to key images they select for each case. Upon logging in, a referring physician can see the report and the key images at a glance and scroll through other medical records and previous lab and imaging reports sorted by date or body part. 

Using a stripped-down radiology workstation for clinicians didn't fit the bill, said Dr. Bradley Erickson, an associate professor of radiology at Mayo. The modified clinical workstation allows more room for text instead of images and saves screens as well. 

"It's common for clinicians to key up the report and images and arrange them in a workroom before meeting with the patient, so they can retrieve that presentation to view with the patient in an efficient and organized fashion," he said. 

Understanding the customer's ultimate needs, however, requires flexibility. 

"We radiologists tend to think we know everything physicians need to know," Erickson said. "We can diagnose a renal cancer and select the images documenting that. But a surgeon planning an approach requires more images than those that prove the diagnosis. Diagnosis is just a starting point. The imaging application must support the whole episode of care." 

Centers around the country are experimenting with ways to improve reports without compromising radiologists' time: 

- The ultrasound department at Brigham and Women's Hospital in Boston uses structured reporting to allow technologists to do initial work on reports, indicating, for example, whether a scan is negative or has a lesion. 

- University of Wisconsin radiologists flag key images for referring physicians and also create teaching files with them. 

Referring physicians like having images embedded in reports as a visual reference, according to Dr. Bruce Reiner, director of radiology research at the VA Maryland Health Care System. In a study presented at the 2004 Society for Computer Applications in Radiology meeting, Reiner and colleagues asked clinical physicians to compare traditional and structured reports with and without images. While the clinicians showed no statistically significant preference for one reporting style over the other, they uniformly preferred having images available. 

While a more comprehensive radiology report could bolster radiology's image, it's a tough sell all around. Radiologists are already overloaded and suspicious of changes that could increase their workload or introduce unfamiliar tasks. Hospital administrators, who saw immediate savings with speech recognition, are skeptical of anything without an established bottom-line value. Vendors don't necessarily want to put in the time and tweaking necessary without a groundswell of demand. And referring clinicians, having never seen proof that phenomena such as structured reporting can improve patient care, may be indifferent. 

Integration and structure still represent a huge hurdle, according to Cooke. As data flow around the healthcare enterprise, they tend to be reduced to what the most primitive system can handle. 

"The challenge for the industry is how to make the insides of the varying systems consistent," he said. 

For radiologists to have complete patient information at their fingertips in the reading room, all hospital systems must be able to contribute to and interact with some sort of detailed patient record. The workstation, meanwhile, needs to take advantage of technology to simplify radiologists' jobs: measuring areas of interest, logging the measurement in the report automatically, comparing it with priors, and so on. 

Such data would be more meaningful if they were consistent, if a suspected tumor were described the same way at one institution as at another. Consistency would facilitate research and in theory improve the clarity of reports. 

"Radiologists can hide in a prose report, and except for BIRADS, they don't have to give a bottom line of their clinical concern," Reiner said. "You can't hide in structured reporting, and that takes radiologists out of their comfort zone." 

In the face of such inertia, it's hard to picture the day when a report that integrates full clinical details, images, and more might circulate in and out of radiology. Nonetheless, Reiner, whose small hospital department uses traditional dictation and prose reporting with no bells or whistles at all, sees the evolution of the radiology report as central to the specialty's survival. 

"Radiology is being commoditized, with third-party and global outsourcing. The radiology community right now is not looking at changes because we're fat and happy," he said. "We need perceived value and quality to reinvent ourselves. As demand increases, we'll be pressured to adopt new digital technologies and build our reports around specific needs." 

A radiologist sitting at a workstation views an abdominal CT study. Priors, matched slice-for-slice, accompany the patient's history and recent lab results. As the speech recognition reporting process begins, software evaluates the text and provides cues to speed and direct the rest of the report. The radiologist adds notes to the images that best display the pathology. Based on the radiologist's conclusion, the software offers links to similar cases and medical references.

Reinvented radiology reports seek renewed momentum

With three shifts, weekend staff, and portable, offsite, and outpatient units all providing computed radiography services, the radiology department at Texas Children's Hospital has plenty of incentive to standardize and improve the quality of the images it collects. With 70% of its case volume coming from CR, the department has been using PACS to track rejected images and develop training for staff.

With some number crunching, the PACS administrator and applications trainer can determine which technologist is prone to overexpose images, who mispositions infants for chest studies, and who uses too much contrast. The department sets a benchmark reject rate and works with individual technologists on the skills they need to hit that mark.

Such projects are one of the often overlooked advantages of PACS: the ability to collate data that previously took more time than most departments had available and to turn it into meaningful quality improvement feedback. That's only the beginning. By its nature, an informatics-driven department also can minimize the risk of clerical errors and lost films; efficiency is a major factor in many quality improvement endeavors. As hospital systems become more thoroughly linked, performance feedback could influence referring physicians' ordering patterns as well.

Updating departments can cause as many errors as it solves. A filmless department no longer needs film librarians to track down missing jackets throughout the hospital. Instead, it needs PACS administrators to reconcile studies for Fry and Frye, who are the same patient with one incorrectly spelled name. Technologists can crop digital images to focus attention on the target anatomy, but in the process they might inadvertently edit out significant peripheral findings. A digital department can be more efficient, but only if the right systems are in place.

At Texas Children's, rejected CR images get a new lease on life. By placing "None" in front of the name, identification number, and other fields, the rejecting technologist or radiologist guarantees that the case will fail to be validated by the radiology information system and will attract the attention of a PACS analyst. By keeping some record that the images were acquired, the department can also track the cumulative radiation exposure for individual patients.

Working with daily reports from PACS analysts and the reject archives, department applications trainers plan semiannual in-services to go over the most common errors department-wide (see table). Team leaders look at each technologist's reject files every month and work individually on needed skills. Some supervisors use this as a learning tool for new employees' 90-day probation period.

"We set a benchmark of a 4.36% reject rate for all CR technologists," said Stephanie Carr, the department applications trainer. "You can see that a new employee might be having problems with positioning. We work with pediatric patients, so there's a certain amount of movement, and the new employee needs to learn techniques to overcome that."

For now the department is limiting its efforts to CR, but Carr and PACS manager Melissa Blado have considered moving the tools to MR and CT once the process is more standardized. Other centers have focused on CR, too, and suggest that it may be more vulnerable to error than other modalities.

"Traditionally, other modalities have been digital from the start," said Ellen Charkot, a technologist at Toronto's Hospital for Sick Children, who promoted the idea of a radiology quality specialist at the 2004 Society for Computer Applications in Radiology meeting. "These technologists were trained on analog and aren't necessarily educated about how CR or DR are created. Things look different, and it takes time to learn how to produce image quality that is consistent with what one is used to seeing. After we started CR in Toronto, for a while the radiologists wondered why every patient looked like they had interstitial pneumonia."

Beyond positioning and exposure errors, the system is as vulnerable as plain film to misidentified or misfiled studies. PACS administrators make manual corrections daily from technologist and radiologist alerts to ensure that what ultimately ends up in the PACS matches what was requested on the RIS.

Although Texas Children's reject system has been in use for several years, it is improvised within the limitations of a four-year-old PACS and an older CR system. Newer software packages allow technologists or radiologists to designate images for teaching files or make sticky notes on screen about specific problems or questions. They also include tools such as modality work lists that reduce the odds that a patient will be misidentified and the file incorrectly archived.

Such problems are serious and time-consuming enough that one healthcare system's radiology department set up a zero-tolerance policy. Tom Hasley, now a digital solutions advisor with Fuji Medical Systems, developed an error reporting and feedback mechanism during his years as a PACS analyst at Rex Healthcare in Raleigh, NC. Errors included images sent to PACS with the wrong accession number, the wrong name, or under the wrong exam title, as well as the wrong patient selected from the work list and sent to PACS.

Technologists who made errors received the following actions:

- second offense, written warning and review training;

- third offense, level 2 written warning and further review training; and

- fourth offense and beyond, disciplinary action and/or termination.

If employees go six weeks without further violations following a verbal warning, their record is cleared unless a pattern of repeat errors develops.

In a department that has 150,000 cases sent to PACS each year, the project rolled out slowly from modality to modality. Between 1999 and 2003, only one employee tested the limits of the program and was fired, Hasley said. The rate of errors dropped, and the system is still in place.

It's a model Dr. Paul Nagy is looking to put in place at the Medical College of Wisconsin, although he plans to take more of a carrot-and-stick approach than one that is strictly punitive.

"Technology isolates the user groups, so it's important to build communication skills," he said. "Radiologists depend on technologists, and technologists need to learn and get feedback. That system needs to be built in and not slow down either party. Radiologists can get frustrated and stop reporting problems if they don't hear that the problems are being addressed."

At Wisconsin, a radiologist who sees a problem with an annotation or technique can right-click it and save a pointer next to it, then put the study in a public area of the PACS with a note on what's wrong. A technologist supervisor in each section checks the reject folder throughout the day and resolves each issue with the appropriate technologist. A logging feature sends an e-mail to the original radiologist with a note that the problem has been addressed.

Such feedback mechanisms are only an addition to actual interpersonal relationships, not a substitute for them, Nagy said. Collaboration and education need to be part of the interaction between radiologist and technologist in the first place. That's easier in interventional radiology and ultrasound, which are collaborative by nature, but radiography can be fairly isolated.

The stakes are higher in informatics environments, because PACS is unforgiving of human error. Missing medical record numbers, patient names, or procedure types can render studies unviewable. Simple steps such as selecting a patient name from the RIS rather than typing it into the study can reduce transcription errors.

As radiology informatics evolves, the gaps for error are closing. A RIS can pull patient demographic data from hospital information systems when a referring physician orders a study. At the modality, the technologist can select the patient's name from a work list and have that data automatically placed into the appropriate fields of the image file. If an identification error does occur, many modalities allow the user to correct information, and newer RIS and PACS should offer options to repair errors as well. Such points are important considerations when purchasing new equipment, said Dr. David Channin, chief of imaging informatics at Northwestern University.

Below the surface, Integrating the Healthcare Enterprise standardization is trying to eliminate nonhuman glitches.

"You have to make sure that natural checks in human systems are supported by automated ones," Channin said. "You need to not only send a file from the CT workstation to PACS, but confirm that PACS got and stored it properly before the CT station can delete it."

Adding consistency to the hardware end and to the human side of radiology holds even greater promise for improving healthcare quality, according to Channin. More detailed clinical and demographic data pulled through the RIS will give radiologists better clues about what to look for in images. Structured reporting will provide referring physicians with more feedback about diagnostic probabilities. Standardizing radiology's language will allow even greater data mining from PACS for research. The RSNA is working on RADLEX, a defined set of terms that can be used in conjunction with SNOMED, the general medical lexicon.

"These lexicons need to define ordering information as well as clinical terms. Something that is 'chest two views' at one facility may be 'x-ray chest PA and LAT' at another," Channin said. "After we get the terms fixed, we can move to develop a true ontology, which defines relationships in a lexicon. 'CT of the trunk' could be universally understood to be CT of the chest, pelvis, and abdomen."

As their products expand from the radiology department throughout the hospital, vendors have embraced the idea that feedback and data-gathering tools can be used to make healthcare more consistent.

"The Holy Grail is a clinical best practices guideline available at the point of order that respects the physician's judgment," said Jess Edwards, PACS product line manager at Eastman Kodak. "You want to help someone make the right decision about whether to order a plain chest film or a body CT."

A sophisticated RIS could incorporate those features, with decision support based on appropriateness criteria, and it could even learn from common ordering patterns at a facility over time. Referring physicians and radiologists would be free to use their best judgment about which study to order, but the preferred workup would be available to them, and hospital quality managers could look at patterns for education detailing or outcomes research.

For the moment, informatics and quality are merging in much less sweeping ways in places like Texas and North Carolina. The data generated with each rejected film let managers benchmark their own performance and customize education. Blado, in particular, is looking forward to a day when she can compare her internal data at Texas Children's Hospital with those from colleagues around the country to see how her department's efforts stack up.

"If you find someone else who's doing this let me know," she said.

The following are the most common causes of rejected CR studies at Texas Children's Hospital.

Appropriateness: Wrong exam performed 3.27%