Dose-volume-based evaluation of convolutional neural network-based auto-segmentation of thoracic organs at risk

Dosimetric impact of contour editing on CT and MRI deep‐learning autosegmentation for brain OARs

“…5 Accordingly, there is little research on the dosimetric effects of contour variations between manual and autosegmentation, and even less on the dosimetric consequences of editing contours either before model training (as here) or post autosegmentation. 6 Recent research 2,3 raises questions about the correlation between common geometric measures, dose planning statistics, and clinical acceptability of OAR contours. Hence, it is difficult to establish whether a segmentation model is clinically usable in a specific clinical scenario, sufficiently limiting the risk of overexposing normal tissue and allowing the precise delivery of RT dose to targets.…”

“…This work is built upon a geometric evaluation which was previously published and hence focusses on the clinically relevant dosimetric aspects. 7 The correlation of dosimetry with the geometric accuracy of MRI and CT-based DL-AC models, established previously, 6,8,9 is also addressed. Further, we determine the dosimetric impact of editing clinical contours to gold standard quality before training CT and MRI DL-AC models.…”

“…6,8,9 Correlations between the geometric and dosimetric measures in thoracic and head and neck OARs have not been identified. 6,8 In contrast, a study investigating esophageal OARs revealed that DSC and OAR dose had a statistically significant overall correlation, although this correlation was not always present at the level of individual patients or OARs. 9 Finally, it is essential to consider the clinical significance of a dosimetric error.…”

Towards real-time radiotherapy planning: The role of autonomous treatment strategies

“…The one-day workflow proposed by Palacios et al [2] used automation only to a minor extent and it is therefore highly dependent on the availability of staff throughout the day and not easily scalable to increasing patient numbers. Automated tools for various steps in the radiotherapy planning workflow such as automatic contouring [4] , [5] , [6] , [7] , [8] and radiotherapy planning [9] , [10] , [11] , [12] , [13] recently gained attention. For instance, Johnston et al [7] showed the usability of a convolutional neural network for segmentation of thoracic organs at risk.…”

“…Automated tools for various steps in the radiotherapy planning workflow such as automatic contouring [4] , [5] , [6] , [7] , [8] and radiotherapy planning [9] , [10] , [11] , [12] , [13] recently gained attention. For instance, Johnston et al [7] showed the usability of a convolutional neural network for segmentation of thoracic organs at risk. Although auto-contouring of targets is more challenging, Xie et al [8] recently introduced a 3D neural network for lung lesion contouring.…”