Comparison of prostate MRI lesion segmentation agreement between multiple radiologists and a fully automatic deep learning system

• Verfasst von: Schelb, Patrick [VerfasserIn]
• Verfasst von: Tavakoli, Andrej [VerfasserIn]
• Verfasst von: Tubtawee, Teeravut [VerfasserIn]
• Verfasst von: Hielscher, Thomas [VerfasserIn]
• Verfasst von: Radtke, Jan Philipp [VerfasserIn]
• Verfasst von: Görtz, Magdalena [VerfasserIn]
• Verfasst von: Schütz, Viktoria [VerfasserIn]
• Verfasst von: Kuder, Tristan Anselm [VerfasserIn]
• Verfasst von: Schimmöller, Lars [VerfasserIn]
• Verfasst von: Stenzinger, Albrecht [VerfasserIn]
• Verfasst von: Hohenfellner, Markus [VerfasserIn]
• Verfasst von: Schlemmer, Heinz-Peter [VerfasserIn]
• Verfasst von: Bonekamp, David [VerfasserIn]
• Titel: Comparison of prostate MRI lesion segmentation agreement between multiple radiologists and a fully automatic deep learning system
• Verf.angabe: Patrick Schelb, Anoshirwan Andrej Tavakoli, Teeravut Tubtawee, Thomas Hielscher, Jan-Philipp Radtke, Magdalena Görtz, Viktoria Schütz, Tristan Anselm Kuder, Lars Schimmöller, Albrecht Stenzinger, Markus Hohenfellner, Heinz-Peter Schlemmer, David Bonekamp
• Jahr: 2021
• Umfang: 15 S.
• Fussnoten: Online veröffentlicht: 19.11.2020 ; Gesehen am 22.03.2024
• Weitere Titel: Abweichender Titel: Zusammenfassung auf Deutsch und Englisch
• Schrift/Sprache: Zusammenfassung unter dem Titel: Vergleich der Kongruenz von Prostata-MRT-Läsionssegmentationen durch mehrere Radiologen und ein vollautomatisches Deep-Learning-System
• Titel Quelle: Enthalten in: RöFo
• Ort Quelle: Stuttgart [u.a.] : Thieme, 1975
• Jahr Quelle: 2021
• Band/Heft Quelle: 193(2021), 5, Seite 559-573
• ISSN Quelle: 1438-9010
• DOI: doi:10.1055/a-1290-8070
• Datenträger: Online-Ressource
• Sprache: eng
• K10plus-PPN: 188407846X

Abstract

<p> <b>Purpose</b> A recently developed deep learning model (U-Net) approximated the clinical performance of radiologists in the prediction of clinically significant prostate cancer (sPC) from prostate MRI. Here, we compare the agreement between lesion segmentations by U-Net with manual lesion segmentations performed by different radiologists.</p> <p> <b>Materials and Methods</b> 165 patients with suspicion for sPC underwent targeted and systematic fusion biopsy following 3 Tesla multiparametric MRI (mpMRI). Five sets of segmentations were generated retrospectively: segmentations of clinical lesions, independent segmentations by three radiologists, and fully automated bi-parametric U-Net segmentations. Per-lesion agreement was calculated for each rater by averaging Dice coefficients with all overlapping lesions from other raters. Agreement was compared using descriptive statistics and linear mixed models.</p> <p> <b>Results</b> The mean Dice coefficient for manual segmentations showed only moderate agreement at 0.48-0.52, reflecting the difficult visual task of determining the outline of otherwise jointly detected lesions. U-net segmentations were significantly smaller than manual segmentations (p < 0.0001) and exhibited a lower mean Dice coefficient of 0.22, which was significantly lower compared to manual segmentations (all p < 0.0001). These differences remained after correction for lesion size and were unaffected between sPC and non-sPC lesions and between peripheral and transition zone lesions.</p> <p> <b>Conclusion</b> Knowledge of the order of agreement of manual segmentations of different radiologists is important to set the expectation value for artificial intelligence (AI) systems in the task of prostate MRI lesion segmentation. Perfect agreement (Dice coefficient of one) should not be expected for AI. Lower Dice coefficients of U-Net compared to manual segmentations are only partially explained by smaller segmentation sizes and may result from a focus on the lesion core and a small relative lesion center shift. Although it is primarily important that AI detects sPC correctly, the Dice coefficient for overlapping lesions from multiple raters can be used as a secondary measure for segmentation quality in future studies.</p> <p> <b>Key Points:</b> </p> <p> <b>Citation Format</b> </p>