AbsCellUSNet - Deep Learning for Detecting Abnormalities in Point-of-Care Ultrasound Maxillofacial Images

Abstract:

Pediatric patients arrive at emergency department with maxillofacial pain, infrequently. They usually come in with a swelling, which is either a cellulitis or abscess. Following the preliminary data collection during patient intake, the first line of diagnosis is using a portable ultrasound setup, also called point-of-care ultrasound system (POCUS). This portable hand-held device makes screening easy and efficient, for the young population, who may be non-cooperative due to acute pain. Abscess in the lower jaw can be dangerous and in the upper jaw it can leak into the orbit, which is equally alarming, and hence needs to be drained immediately before prescriptive treatment. Whereas cellulitis may or may not require draining and can be treated with antibiotics in most cases. POCUS is also used as a post-operative measure to confirm the infectious swellings have successfully drained. Addressing the need for improved diagnostic tools, we compared AI models to obtain valuable insights for applying deep learning to POCUS image analysis in this clinical domain. 

Example Classification Results with AbsCellUSNet:

TBA

References:

[1] V. B. S. Prasath et al. Improving oromaxillofacial infection diagnosis in children with handheld point-of-care ultrasound and deep learning. Submitted, 2025.

[3] V. B. S. Prasath et al. Anonymous submission. Submitted, 2025.

Back to Ultrasound Projects Main Page. Back to Research Page.