Free Air Detection

Pneumoperitoneum refers to the presence of air in the peritoneal space (the abdomen). The most common cause of pneumoperitoneum is gastrointestinal perforation. Patients with this disorder have a high 30-day mortality of up to 31% with early surgical treatment being crucial for survival. This puts forward the idea that early detection of pneumoperitoneum could help reduce the mortality of patients with gastrointestinal perforation. An AI algorithm, which could detect pneumoperitoneum, would have a direct positive impact on patient survival. The disease fortunately is quite rare and found in less than 1% of all scans. That means that false positive findings needed to be avoided in order to reduce alert fatigue for the radiologist.

Together with Siemens Healthineers we designed an algorithm which could detect small free air bubbles in the abdominal cavity with a very high specificity (>99%) at a moderate sensitivity. [Ref 1] Trained on just over 100 mostly postoperative patients, the network performed well. We tested this algorithm in a retrospective clinical validation in a consecutive cohort and found a similar high specificity at moderate sensitivity. [Ref2]”

References

[Ref 1] Taubmann, O. et al. (2020). Automatic Detection of Free Intra-abdominal Air in Computed Tomography. In: , et al. Medical Image Computing and Computer Assisted Intervention – MICCAI 2020. MICCAI 2020. Lecture Notes in Computer Science(), vol 12262. Springer, Cham.

[Ref2] Brejnebøl, M. et al (2022). Artificial Intelligence based detection of pneumoperitoneum on CT scans in patients presenting with acute abdominal pain: A clinical diagnostic test accuracy study. European Journal of Radiology.