Precision of manual two-dimensional segmentations of lung and liver metastases and its impact on tumour response assessment using RECIST 1.1

Abstract

Background: Response evaluation criteria in solid tumours (RECIST) has significant limitations in terms of variability and reproducibility, which may not be independent. The aim of the study was to evaluate the precision of manual bi-dimensional segmentation of lung, liver metastases, and to quantify the uncertainty in tumour response assessment. Methods: A total of 520 segmentations of metastases from six livers and seven lungs were independently performed by ten physicians and ten scientists on CT images, reflecting the variability encountered in clinical practice. Operators manually contoured the tumours, firstly independently according to the RECIST and secondly on a preselected slice. Diameters and areas were extracted from the segmentations. Mean standard deviations were used to build regression models and 95% confidence intervals (95% CI) were calculated for each tumour size and for limits of progressive disease (PD) and partial response (PR) derived from RECIST 1.1. Results: Thirteen aberrant segmentations (2.5%) were observed without significant differences between the physicians and scientists; only the mean area of liver tumours (p = 0.034) and mean diameter of lung tumours (p = 0.021) differed significantly. No difference was observed between the methods. Inter-observer agreement was excellent (intra-class correlation >0.90) for all variables. In liver, overlaps of the 95% CI with the 95% CI of limits of PD or PR were observed for diameters above 22.7 and 37.9 mm, respectively. An overlap of 95% CIs was systematically observed for area. No overlaps were observed in lung. Conclusions: Although the experience of readers might not affect the precision of segmentation in lung and liver, the results of manual segmentation performed for tumour response assessment remain uncertain for large liver metastases.

Radiologists and Clinical Trials: Part 1 The Truth About Reader Disagreements

“…Therefore, measurement differences among readers due to differences in the determination of the measured edge or the longest axis can affect the target lesion response. [ 81 ] Measurement variability between readers for the same lesion, for most tumor types, can be considered to be only a minor contributor to overall reader variability (intraclass correlation coefficient = 0.991) [ 82 ].…”

Deep learning and level set approach for liver and tumor segmentation from CT scans

“…Liver surgical planning treatments would benefit from an accurate and fast liver and tumor segmentation that allows for subsequent determination of tumor burden and texture‐based information. Moreover, having a standardized and automatic segmentation method would facilitate a more reliable therapy response classification 7 …”

“…Moreover, having a standardized and automatic segmentation method would facilitate a more reliable therapy response classification. 7 Organ segmentation from CT scans has been a hot research topic during the past few years. Recently, due to the advancement in computer vision, the development of deep fully convolutional neural (FCNs) networks enhanced the performance of the semantic segmentation, and leads to outperform other competitors in the field of medical imaging.…”

Automatic liver tumor segmentation in CT with fully convolutional neural networks and object-based postprocessing

“…Liver therapy planning procedures would profit from an accurate and fast lesion segmentation that allows for subsequent determination of volume- and texture-based information. Moreover, having a standardized and automatic segmentation method would facilitate a more reliable therapy response classification 2 .…”