In this video toolkit for ventral hernia repair, you’ll see experienced surgeons perform a range of techniques using SurgiMend®, a biologically derived matrix.
Imagine if you could input hernia patients’ preoperative scans into a computer program and have artificial intelligence predict, with a significant degree of accuracy, which patients will need component separation surgery, develop a post-surgical wound infection or suffer from post-surgical pulmonary failure?
Not only could this data help surgeons determine which hernia repair patients need specialty care, but it might also determine whether a patient proceeds with surgery once they can see the risks involved.
That day is closer than ever. Hernia specialists at Atrium Health Carolina Medical Center have developed what they say is the world’s first computer program that uses only objective and easily accessible patient data to make predictions about hernia surgery complexity. The program makes predictions based on a hernia repair patient’s preoperative CT scan.
In testing, the computer program was:
- 89% accurate at identifying patients who would develop a post-surgical wound infection
- 75% accurate at identifying patients who would require component separation
- 54.5% accurate in predicting pulmonary failure
“Almost every complex hernia patient gets a CT scan,” Dr. B. Todd Heniford, chief of the Division of Gastrointestinal and Minimally Invasive Surgery at Atrium Health, said. “We thought if we could base our predictions on that imagery alone, we’d have a tool that almost any surgeon in the world could use to get a clear picture of a patient’s needs.”
Training the Predictive Model
Dr. Heniford partnered with Drs. Vedra Augenstein, Paul Colavita, Kent Kercher, and general surgery resident Dr. Sharbel Elhage, all at Atrium Health. First, they created layers of algorithmic calculations, called a neural network, to form the artificially intelligent “brain.”
Then they trained the program by inputting CT scans from hundreds of past patients. The program reviewed each image more than 12,000 times, recording features that corresponded with the need for component separation, the development of wound infection and post-surgical pulmonary failure.
The team then gave the program another set of images from other past patients and again asked it to predict the three outcomes of repairing a hernia. Next, the team asked an international panel of hernia repair experts to review the same scans and predict which patients would require component separation.
The results? The AI was 75% accurate at predicting surgical complexity in hernia repair, nearly 15 points better than the experts.
Building on Innovation
This isn’t the first predictive tool that Dr. Heniford and his team at Atrium Health have developed. In 2014, they created CeDAR, a free app available for download on smartphone. The app asks eight health related questions, including those about diabetes, smoking and previous hernia repair.
The app then uses the responses to calculate the patient’s risk of post-surgical complication for treatment of a ventral hernia. It also provides estimated costs for in-hospital and follow-up care in excess of the national average for treatment in various hernia repair options.
While many hernias are straightforward, up to 30% of patients will develop complications such as wound infections or pulmonary failure. A similar percentage of patients will have a hernia recurrence, often because of wound complications or because the abdomen wasn’t completely closed during the initial surgery. Comorbidities, including smoking, obesity and uncontrolled diabetes, are associated with higher recurrence and complication rates. Risk of recurrence after a hernia repair lasts a lifetime.
The tool can benefit both surgeons and patients. A patient could use the tool to make a decision about whether or not to have hernia surgery, and if so, whether the potential complications would necessitate surgery at a specialty center. For example, repairing large or complex hernias often requires using an advanced technique such as component separation.
Heniford’s team plans to share their artificial intelligence invention with other institutions in order to test it using images of 100 different patients. They then plan to turn the technology into another app so hernia surgeons around the world will have access to this predictive tool.
“Imagine,” Dr. Heniford said, “dragging an image into this program and getting three clear answers: Yes, this patient needs component separation; yes, they’ll have an infection; no, they won’t have a lung problem. That is game-changing clarity.”
The team hopes this technology may someday help predict complexity in other surgeries, including liver and lung tumors.
Read more about the team’s work in the Journal of the American Medical Association.