Algorithmic approach 90% accurate at diagnosing pediatric acute appendicitis

Researchers in Germany have developed an AI-based method for identifying pediatric acute appendicitis using biomarkers like blood and protein readings obtained in routine lab tests.

Their technique is “capable of becoming the gold standard” for confirming or ruling out the diagnostically tricky condition, according to a study report published in PLOS One, and it also proved capable of differentiating between complicated and uncomplicated appendicitis.

The latter task is a prerequisite for guiding appropriate treatment, note the authors, who were led by Marc Reismann, PhD, head of the pediatric surgical research department at Charité academic hospital in Berlin.

“Appendectomy is still the therapy of choice, but conservative strategies are increasingly being studied for uncomplicated inflammation,” the authors explain.

The team trained AI algorithms on full blood counts, c-reactive protein (CRP) and appendiceal diameters in abdominal ultrasound exams of 590 children and adolescents. Of these, 473 had confirmed appendicitis and 117 were negative for the illness.

Retrospectively analyzing the diagnostic performance of their AI algorithms, the researchers found its accuracy reached 90% (93% sensitivity, 67% specificity).

The tool was successful more than half the time at correctly identifying complicated versus uncomplicated inflammation.

“Such a test would be capable to prevent two out of three patients without appendicitis from useless surgery as well as one out of three patients with uncomplicated appendicitis,” Reismann et al. write.

“The presented method has the potential to change today’s therapeutic approach for appendicitis,” the authors comment, adding that it demonstrates “the capability of algorithms from artificial intelligence and machine learning to significantly improve diagnostics even based on routine diagnostic parameters.”

The study is available in full for free.