New Technology: Using AI To Interpret Pelvic X-rays

October 23, 2023 The Trauma Pro Leave a comment

Look out, radiologists! The computers are coming for you!

Radiologists use their extensive understanding of human anatomy and combine it with subtle findings they see on x-ray shadow pictures. In doing this, they can identify a wide variety of diseases, anomalies, and injuries. But as we have seen with vision systems and game playing (think chess), computers are getting pretty good at doing this as well.

Is it only a matter of time until computer artificial intelligence (AI) starts reading x-rays? Look at how good they already are at interpreting EKGs. The trauma group at Stanford paired up with the Chang Gung Memorial Hospital in Taiwan to test the use of AI for interpreting images to identify a specific set of common pelvic fractures.

The Stanford group used a deep learning neural network (XCeption) to analyze source x-rays (standard A-P pelvis images) from Chang Gung. These x-rays were divided into training and testing cohorts. The authors also applied different degrees of blurring, brightness, rotation, and contrast adjustment to the training set in order to help the AI overcome these issues when interpreting novel images.

The AI interpreted the test images with a very high degree of sensitivity, specificity, accuracy, and predictive values, with all of them over 0.90. The algorithms generated a “heat map” that showed the areas that were suspicious for fracture. Here are some examples with the original x-ray on the left and the heat map on the right:

The top row shows a femoral neck fracture, the middle row an intertrochanteric fracture, and the bottom row another femoral neck fracture with a contralateral implant. All were handily identified by the AI.

AI applications are usually only as good as their training sets. In general, the bigger the better so they can gain a broader experience for more accurate interpretation. So it is possible that uncommon, subtle fractures could be missed. But remember, artificial intelligence is meant to supplement the radiologist, not replace him or her. You can all breathe more easily now.

This technology has the potential for broader use in radiographic interpretation. In my mind, the best way to use it is to first let the radiologist read the images as they usually do. Once they have done this, then turn on the heat map so they can see any additional anomalies the AI has found. They can then use this information to supplement the initial interpretation.

Expect to see more work like this in the future. I predict that, ultimately, the picture archiving and communications systems (PACS) software providers will build this into their product. As the digital images are moving from the imaging hardware to the digital storage media, the AI can intercept it and begin the augmented interpretation process. The radiologist will then be able to turn on the heat map as soon as the images arrive on their workstation.

Stay tuned! I’m sure there is more like this to come!

Reference: Practical computer vision application to detect hip fractures on pelvic X-rays: a bi-institutional study. Trauma Surgery and Acute Care Open 6(1), http://dx.doi.org/10.1136/tsaco-2021-000705.

What the heck?

What’s With Those John / Jane Doe Names?

October 10, 2023 The Trauma Pro Leave a comment

Oftentimes, trauma patients arrive, are not very responsive, and are not carrying an ID. However, our electronic health record systems have a very difficult time with this. To expedite care, most hospitals developed a system of pseudo-names to be used until the patient can be identified.

Originally, these names were often “Doe” names. The classic ones were John Doe and Jane Doe. Obviously, more than two names were needed, so other first names were adopted to provide a sizable pool of pseudo-names. Other disciplines, such as law enforcement, have used a similar system.

Where did the concept of John/Jane Doe come from? The origin is the British legal system, way back in the 18th century. In those days, landlords would start a “process of ejectment,” known these days as eviction, to rid their properties of landlords or deadbeat tenants. The legal process was a bit complex, so they would file for the process using a fake name to initiate it. They frequently selected John Doe or Richard Roe for their filings.

The reason the Doe and Roe names were used has been lost to history. There is speculation that these names were derived from certain deer species endemic to Britain. But honestly, no one knows for sure now. Although this process was dropped in Great Britain in the mid-1800s, it persists in the US legal system. Remember the landmark legal case Roe v Wade?

The use of Doe names in electronic health records is rapidly fading as well. The list of name pairs ending with Doe is just not unique enough. There are too many opportunities to mix up similar names, especially if the last one is always the same. This can result in catastrophic errors if test results are misinterpreted, or a blood transfusion with incorrect ABO typing is given to the wrong patient.

Most trauma centers have adopted temporary naming systems consisting of two words or unique names. Some use number and letter sequences combined with another unique word. The real trick in the electronic medical record world is smoothly merging the records utilizing the pseudo-name with the patient’s previous records under their actual name.

The best practice for this varies by electronic record system and hospital. If done too early, the change may disrupt critical processes, such as the massive transfusion protocol. If done too late, it is difficult for trauma professionals to see any records the patient may have under their actual name. Each center must develop its own system for converting from John Doe to the real name.

Complications, Thorax

Early Vs Delayed Thoracic Endovascular Repair

October 6, 2023 The Trauma Pro Leave a comment

Back in the day, the only way to fix a broken thoracic aorta was via left thoracotomy. This was a big procedure, with the possibility of several major complications, with postop paraplegia being one of them. At the time, there was a debate about whether the procedure should be done immediately versus waiting until the patient was well-resuscitated. The concern was that death was nearly certain if the aortic lesion progressed.

We learned that temporizing with strict blood pressure control worked wonders at protecting the patient. Although many of these injuries were managed within hours, a growing number were delayed by a few days to improve outcomes.

Nowadays, thoracic endovascular aortic repair (TEVAR) is routine and much less morbid than the open procedure. However, the same question arises: do it early or wait a while? Interestingly, not one but two analyses have been published on this very topic in the last four months!

The first is from an international research group that searched the usual databases and initially found 921 records. They included only clinical trials or cohort studies with ten or more adult patients that could be stratified as early (within 24 hours) or late (after 24 hours) intervention. After applying these criteria, only seven studies remained for analysis.

There were 3,757 patients with early repairs, compared to 1,238 undergoing late repair. The presenting demographics and injury grades were similar in each group. However, the short-term mortality was significantly higher (1.9x) in the early TEVAR group. Additionally, ICU length of stay was significantly longer (3 days) in the late TEVAR group.

The second paper was presented as a quick-shot at last year’s AAST meeting. It is from a group of researchers from our big Boston trauma centers and the Netherlands. They used four years of data from the TQIP database, giving them extra information unavailable in the first study. They specifically looked at patients with grade II or III injuries. Here is the grading scale:

Here are the factoids:

A total of 1,339 patients were studied, with about three-quarters in the early TEVAR group
Median time to TEVAR was 4 hours in the early group and 65 hours in the late group
Patients in the early group were significantly less likely to have brain or liver injuries
ISS was similar in both groups
The early TEVAR group had significantly higher in-hospital mortality (16% vs. 5%), significantly higher risk of ARDS (7.6% vs. 2.1%), but significantly shorter ICU stay (7 vs 10 days)
When patients who died within the first 24 hours were excluded, the in-hospital mortality remained significantly higher, and the ICU and hospital lengths of stay were significantly shorter

Bottom line: Some society guidelines began recommending delayed TEVAR in 2015. This study did not detect any trend toward this, however. Using different methods and databases, these two studies identified nearly identical mortality and ICU trends in large groups of patients. The mortality trends do not appear to be related to injury grade, overall injury severity, or the presence of head injury.

Taken together, this suggests that we need to rethink the timing of TEVAR in patients with grade II or III injuries. The best timing still needs to be defined, but it appears to be beyond 24 hours. Centers performing this procedure should review their results and consider extending procedure timing as additional research is done to define the ideal time interval.

References:

Early Versus Delayed Thoracic Endovascular Aortic Repair for Blunt Traumatic Aortic Injury: A Systematic Review and Meta-Analysis. Cureus. 2023 Jun 28;15(6):e41078. doi: 10.7759/cureus.41078. PMID: 37519486; PMCID: PMC10375940.
Early Versus Delayed Thoracic Endovascular Aortic Repair for
Blunt Thoracic Aortic Injury: A Propensity Score-Matched Analysis. Ann Surg 278:e848-e854, 2023.

Complications

Back Braces: Are They Really Needed?

October 3, 2023 The Trauma Pro Leave a comment

Back braces have always confused me. There are so many types: TLSO, LSO, backpack, extension, and even the lowly abdominal binder can function as a brace. And I have never been able to predict which brace my spine colleagues accurately would prescribe for a specific condition or injury.

Many vertebral fractures can be treated non-operatively. And it seems intuitive that there would be some benefit from splinting the spine to limit the range of motion to enhance healing. So I would like to concentrate on some papers that examined the use of back braces on patients who underwent pedicle screw fixation of their thoracic and/or lumbar spine fractures.

I found two systematic reviews and a ten-year prospective clinical trial. First, the reviews.

The first was published in 2016 and examined papers comparing postop bracing vs. no postop bracing. It looked at loss of deformity correction, return to work, functional improvement, instrumentation failure rate, pseudoarthrosis, and postop complications. A total of 76 studies were included.

The average wear time for the braces was just over three months. There was no difference in pain, return to work, functional outcome, or instrumentation failure. Interestingly, there was a significant increase in the number of patients who lost kyphotic reduction from the procedure and a significant increase in the number of complications. Unfortunately, the authors did not break out specific complications other than wound infections. Although these were higher in the braced group (2.8% vs. 1.8%), this difference was not significant. One real positive for the brace: the number of pseudoarthroses was significantly less (2.4% vs. 6%).

The next review looked at postoperative bracing from a cost-effectiveness standpoint. It considered adverse events such as infections and hardware failure and examined a total of 1957 patients across 48 papers. Non-braced patients were somewhat older. Braced patients had significantly fewer reoperations for non-union or hardware failure (1.3% vs. 1.8%). There was no difference in wound dehiscence or infection. Overall, there was no cost benefit to applying a back brace in these patients after pedicle screw fixation.

The last paper was a prospective clinical trial that enrolled 144 patients randomized to brace or no brace after their surgical procedure. They underwent multiple postop evaluations to monitor quality of life, success of the fusion, pain, mobility, and return to previous activities.

Here are the factoids:

Mean age was only 34 years, and only fractures of T11 to L5 were included
Two-thirds of the injuries were due to falls, and most of the remaining ones were from car crashes
Three-quarters were burst fractures, and the remaining were wedge or fracture-dislocation injuries
There were no differences in early mobilization, residual pain, return to previous activities, quality of life, or success of the fusion after one year

The authors concluded that the use (or non-use) of a TLSO brace after pedicle screw fixation of these fractures does not affect treatment. They state that TLSO braces are not required in the postop period for these patients.

Bottom line: Interesting, yet slightly confusing. The systematic reviews show a significant increase in pseudoarthrosis in patients without braces. This was not seen in the prospective study. The discrepancy could be due to the quality of the papers included in the reviews or insufficient power in the prospective study. So it may be too early to fully know the difference. But it appears to be so small that the question of whether a brace is really necessary needs to be asked.

Most braces are expensive and uncomfortable. I have seen several patients in follow-up who basically stopped wearing their brace as soon as they were out of the hospital. And anecdotally, they did fine.

A large, multicenter study in progress in the Netherlands will analyze the same factors listed above. Unfortunately, the listing on clinicaltrials.gov indicates they are only recruiting 45 patients, which may not add much to what is already known due to low statistical power.

In the meantime, consider discussing with your spine surgeons reviewing the current literature on back bracing. It may be possible to reduce the number of eligible patients who take their brace home and use it as a chew toy for the dog.

References:

Bracing After Surgical Stabilization of Thoracolumbar Fractures: A Systematic Review of Evidence, Indications, and Practices. World Neurosurg 93:221-228, 2016.
Post-operative bracing after pedicle screw fixation for thoracolumbar burst fractures: A cost-effectiveness study. J Clin Neurosci 45:33-39, 2017.
Evaluation of postoperative bracing on unstable traumatic lumbar fractures after pedicle screw fixation. Int J Burns Trauma 12(4):168-174, 2022.
Is postoperative bracing after pedicle screw fixation of spine fractures necessary? Study protocol of the ORNOT study: a randomised controlled multicentre trial. NCT03097081. Listed in 2017, to be completed Nov 2026.

Home of the Trauma Professional's Blog