Family of boundary overlap metrics for the evaluation of medical image segmentation

Abstract

All medical image segmentation algorithms need to be validated and compared, yet no evaluation framework is widely accepted within the imaging community. None of the evaluation metrics that are popular in the literature are consistent in the way they rank segmentation results: they tend to be sensitive to one or another type of segmentation error (size, location, and shape) but no single metric covers all error types. We introduce a family of metrics, with hybrid characteristics. These metrics quantify the similarity or difference of segmented regions by considering their average overlap in fixed-size neighborhoods of points on the boundaries of those regions. Our metrics are more sensitive to combinations of segmentation error types than other metrics in the existing literature. We compare the metric performance on collections of segmentation results sourced from carefully compiled two-dimensional synthetic data and three-dimensional medical images. We show that our metrics: (1) penalize errors successfully, especially those around region boundaries; (2) give a low similarity score when existing metrics disagree, thus avoiding overly inflated scores; and (3) score segmentation results over a wider range of values. We analyze a representative metric from this family and the effect of its free parameter on error sensitivity and running time.

Automated Segmentation of Infarct Lesions in T1-Weighted MRI Scans Using Variational Mode Decomposition and Deep Learning

“…Unlike IoU and DSC, this index is a distance measurement. It calculates the average of all the distances from voxels on the boundary of the segmented lesion to those of the ground truth and vice versa [ 30 ]. The smaller number of ASSD indicates the better segmentation performance.…”

Synthetic medical image generator for data augmentation and anonymisation based on generative adversarial network for glioblastoma tumors growth prediction

“…In this work, we propose an end-to-end method to improve the prediction of GBM growth based on MRI images with several data augmentation techniques. TGP model takes as input MRI image (four channels) with an aim to predict the tumour volume in the late-stage (tumour area after 90 days) in terms of three metrics [45]: recall, precision, and dice scores, each metric is calculated using four statistical values:…”

“…Afterwards, we compare the two output masks in terms of three metrics [45]: recall, precision, and Dice scores. Fig.…”

Automated segmentation of endometrial cancer on MR images using deep learning

“…A low Hausdorff distance indicates high segmentation accuracy in terms of boundary similarity. In order to map segmentation errors in the outer tumor boundaries, we also computed average surface distance (ASD) measured from the ground truth mask surface to the ML mask surface in the Euclidean sense 28 . A smaller value of ASD is associated with an improved segmentation.…”