Fully automated deep learning based auto-contouring of liver segments and spleen on contrast-enhanced CT images

Abstract

Abstract Manual delineation of liver segments on computed tomography (CT) images for primary/secondary liver cancer (LC) patients is time-intensive and prone to inter/intra-observer variability. Therefore, we developed a deep-learning-based model to auto-contour liver segments and spleen on contrast-enhanced CT (CECT) images. We trained two models using 3d patch-based attention U-Net ( $${{\text{M}}}_{{\text{paU}}-{\text{Net}}})$$ M paU - Net ) and 3d full resolution of nnU-Net ( $${{\text{M}}}_{{\text{nnU}}-{\text{Net}}})$$ M nnU - Net ) to determine the best architecture ( $${\text{BA}})$$ BA ) . BA was used with vessels ( $${{\text{M}}}_{{\text{Vess}}})$$ M Vess ) and spleen ( $${{\text{M}}}_{{\text{seg}}+{\text{spleen}}})$$ M seg + spleen ) to assess the impact on segment contouring. Models were trained, validated, and tested on 160 ( $${{\text{C}}}_{{\text{RTTrain}}}$$ C RTTrain ), 40 ( $${{\text{C}}}_{{\text{RTVal}}}$$ C RTVal ), 33 ( $${{\text{C}}}_{{\text{LS}}}$$ C LS ), 25 (CCH) and 20 (CPVE) CECT of LC patients. $${{\text{M}}}_{{\text{nnU}}-{\text{Net}}}$$ M nnU - Net outperformed $${{\text{M}}}_{{\text{paU}}-{\text{Net}}}$$ M paU - Net across all segments with median differences in Dice similarity coefficients (DSC) ranging 0.03–0.05 (p 0.05), however, both were slightly better than $${{\text{M}}}_{{\text{Vess}}}$$ M Vess by DSC up to 0.02. The final model, $${{\text{M}}}_{{\text{seg}}+{\text{spleen}}}$$ M seg + spleen , showed a mean DSC of 0.89, 0.82, 0.88, 0.87, 0.96, and 0.95 for segments 1, 2, 3, 4, 5–8, and spleen, respectively on entire test sets. Qualitatively, more than 85% of cases showed a Likert score $$\ge$$ ≥ 3 on test sets. Our final model provides clinically acceptable contours of liver segments and spleen which are usable in treatment planning.