Convolutional neural network metal artifact reduction in ultra-high resolution photon counting detector CT of middle- and inner-ear implants

Metallic inner- and middle-ear implants with submillimeter structures requiring ultra-high resolution (UHR) imaging invariably result in metal artifacts at CT, which can obscure adjacent microstructures and subtle pathologic findings....

Continue reading at PubMed →

No actionable change yet — findings are a technical proof-of-concept; clinical validation in larger patient cohorts is needed before this AI-based imaging approach should influence implant imaging protocols.

Reliable high-resolution CT imaging of ear implants is a persistent clinical challenge, and AI-based artifact reduction on next-generation photon counting CT scanners could meaningfully improve post-operative assessment of implant positioning.

1. 01A convolutional neural network (AI model) was evaluated for reducing metal-caused imaging errors in CT scans of ear implants.
2. 02The CT system used is an ultra-high resolution photon counting detector — a newer, more detailed scanner than conventional CT.
3. 03Target structures include submillimeter (smaller than 1 mm) middle- and inner-ear implant components.
4. 04Metal artifacts (streaks and blurring from implants) are a known limitation of standard CT post-implantation imaging.
5. 05AI artifact reduction could improve visualization of implant placement and surrounding anatomy.

PMID

42875916

DOI

10.1002/mp.70711.

Journal

Medical Physics

Publication type

research_article

Evidence level

4

Population

CT phantom or cadaveric/clinical images of middle- and inner-ear implants (submillimeter structures)

Intervention

Convolutional neural network metal artifact reduction algorithm applied to ultra-high resolution photon counting detector CT

Comparator

Standard CT reconstruction without AI artifact reduction