Szczegóły publikacji
Opis bibliograficzny
Image segmentation for automatic needle puncture annotation in vibroacoustic signals / Karol Strama, Katharina Steeg, Dominik RZEPKA, Artur KOS, Gabriele A. Krombach, Katarzyna HERYAN, Michael FRIEBE // Current Directions in Biomedical Engineering [Dokument elekroniczny]. — Czasopismo elektroniczne . — 2025 — vol. 11 iss. 1, s. 298–301. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 301, Abstr. — Publikacja dostępna online od: 2025-09-12. — 2025 Joint Annual Conference of the Austrian (ÖGBMT), German (VDE DGBMT) and Swiss (SSBE) Societies for Biomedical Engineering : 9–11 September 2025, Muttenz/Basel, Switzerland
Autorzy (7)
- AGHStrama Karol
- Steeg Katharina
- AGHRzepka Dominik
- AGHKos Artur
- Krombach Gabriele
- AGHHeryan Katarzyna
- AGHFriebe Michael
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 165012 |
|---|---|
| Data dodania do BaDAP | 2025-12-16 |
| Tekst źródłowy | URL |
| DOI | 10.1515/cdbme-2025-0176 |
| Rok publikacji | 2025 |
| Typ publikacji | referat w czasopiśmie |
| Otwarty dostęp |  |
| Creative Commons | |
| Czasopismo/seria | Current Directions in Biomedical Engineering |
Abstract
Manual annotation of biomedical data is often a prolonged and error-prone process, particularly in scenarios involving dynamic physical interactions, such as needle insertions into soft tissue. In this study, we present an automated annotation framework using the Segment Anything Model (SAM) to detect puncture events in vibroacoustic recordings from needle insertions into preserved Manduca sexta specimens. The annotation tool utilizes SAM to extract segmentation masks from video recordings of needle insertions and analyzes vertical deformations of the resulting mask boundaries. Two puncture moments, one at entry and one at exit, were estimated by searching for minima in the bottom boundary displacements, yielding more reliable results compared to the top boundary. The tool detected the first and second punctures in 85% and 90% out of 300 cases, respectively. Although the tool provides promising accuracy, further refinement could be done to achieve higher and more robust precision. Limitations like background interference, reflections, and needle inclusion can be mitigated by improving the setup or optimizing segmentation models, including fine-tuning SAM.
