González, Camila ; Gotkowski, Karol ; Fuchs, Moritz ; Bucher, Andreas ; Dadras, Armin ; Fischbach, Ricarda ; Kaltenborn, Isabel Jasmin ; Mukhopadhyay, Anirban (2022)
Distance-based detection of out-of-distribution silent failures for Covid-19 lung lesion segmentation.
In: Medical Image Analysis, 82
doi: 10.1016/j.media.2022.102596
Artikel, Bibliographie
Kurzbeschreibung (Abstract)
Automatic segmentation of ground glass opacities and consolidations in chest computer tomography (CT) scanscan potentially ease the burden of radiologists during times of high resource utilisation. However, deep learningmodels are not trusted in the clinical routine due to failing silently on out-of-distribution (OOD) data. Wepropose a lightweight OOD detection method that leverages the Mahalanobis distance in the feature space andseamlessly integrates into state-of-the-art segmentation pipelines. The simple approach can even augment pretrainedmodels with clinically relevant uncertainty quantification. We validate our method across four chest CTdistribution shifts and two magnetic resonance imaging applications, namely segmentation of the hippocampusand the prostate. Our results show that the proposed method effectively detects far- and near-OOD samplesacross all explored scenarios.
| Typ des Eintrags: | Artikel |
|---|---|
| Erschienen: | 2022 |
| Autor(en): | González, Camila ; Gotkowski, Karol ; Fuchs, Moritz ; Bucher, Andreas ; Dadras, Armin ; Fischbach, Ricarda ; Kaltenborn, Isabel Jasmin ; Mukhopadhyay, Anirban |
| Art des Eintrags: | Bibliographie |
| Titel: | Distance-based detection of out-of-distribution silent failures for Covid-19 lung lesion segmentation |
| Sprache: | Englisch |
| Publikationsjahr: | November 2022 |
| Verlag: | Elsevier |
| Titel der Zeitschrift, Zeitung oder Schriftenreihe: | Medical Image Analysis |
| Jahrgang/Volume einer Zeitschrift: | 82 |
| DOI: | 10.1016/j.media.2022.102596 |
| Kurzbeschreibung (Abstract): | Automatic segmentation of ground glass opacities and consolidations in chest computer tomography (CT) scanscan potentially ease the burden of radiologists during times of high resource utilisation. However, deep learningmodels are not trusted in the clinical routine due to failing silently on out-of-distribution (OOD) data. Wepropose a lightweight OOD detection method that leverages the Mahalanobis distance in the feature space andseamlessly integrates into state-of-the-art segmentation pipelines. The simple approach can even augment pretrainedmodels with clinically relevant uncertainty quantification. We validate our method across four chest CTdistribution shifts and two magnetic resonance imaging applications, namely segmentation of the hippocampusand the prostate. Our results show that the proposed method effectively detects far- and near-OOD samplesacross all explored scenarios. |
| Freie Schlagworte: | Out-of-distribution detection, Uncertainty estimation, Distribution shift |
| Zusätzliche Informationen: | Art.No.: 102596 |
| Fachbereich(e)/-gebiet(e): | 20 Fachbereich Informatik 20 Fachbereich Informatik > Graphisch-Interaktive Systeme |
| Hinterlegungsdatum: | 17 Feb 2023 07:57 |
| Letzte Änderung: | 10 Jul 2023 15:16 |
| PPN: | 509471323 |
| Export: | |
| Suche nach Titel in: | TUfind oder in Google |
 |
Frage zum Eintrag |
Optionen (nur für Redakteure)
 |
Redaktionelle Details anzeigen |
Drucken
Impressum