Accuracy and practical aspects of semi- and fully automatic segmentation methods for resected brain areas

Abstract

Purpose Precise segmentation of brain lesions is essential for neurological research. Specifically, resection volume estimates can aid in the assessment of residual postoperative tissue, e.g. following surgery for glioma. Furthermore, behavioral lesion-symptom mapping in epilepsy relies on accurate delineation of surgical lesions. We sought to determine whether semi- and fully automatic segmentation methods can be applied to resected brain areas and which approach provides the most accurate and cost-efficient results. Methods We compared a semi-automatic (ITK-SNAP) with a fully automatic (lesion_GNB) method for segmentation of resected brain areas in terms of accuracy with manual segmentation serving as reference. Additionally, we evaluated processing times of all three methods. We used T1w, MRI-data of epilepsy patients (n = 27; 11 m; mean age 39 years, range 16–69) who underwent temporal lobe resections (17 left). Results The semi-automatic approach yielded superior accuracy (p < 0.001) with a median Dice similarity coefficient (mDSC) of 0.78 and a median average Hausdorff distance (maHD) of 0.44 compared with the fully automatic approach (mDSC 0.58, maHD 1.32). There was no significant difference between the median percent volume difference of the two approaches (p > 0.05). Manual segmentation required more human input (30.41 min/subject) and therefore inferring significantly higher costs than semi- (3.27 min/subject) or fully automatic approaches (labor and cost approaching zero). Conclusion Semi-automatic segmentation offers the most accurate results in resected brain areas with a moderate amount of human input, thus representing a viable alternative compared with manual segmentation, especially for studies with large patient cohorts.

Deep learning-based automated segmentation of resection cavities on postsurgical epilepsy MRI

“…This measure characterizes the segmentation border reliability. Several variants of the Hausdorff distance are reported in the literature ( Bakas et al, 2018 , Gau et al, 2020 ); we report the 95 % Hausdorff distance, which is more robust to outliers. Additionally, the relationship between manual and automated resection volumes was plotted for each subject and we report the Pearson’s correlation coefficient and mean absolute error (MAE) between these variables.…”

“…Previous studies have found a relationship between lesion size and classifier accuracy as measured by DSC and percent volume difference (PVD) between predicted and manual segmentations ( Gau et al, 2020 ). To understand whether lesion size contributed to classifier accuracy or PVD error, we partitioned subjects into small (N = 17) and large (N = 28) resection groups using the same threshold (17.92 ml) previously reported ( Tan and EfficientNet, 2019 ).…”

“…One advantage of the present study is epilepsy patient data was used during model training. A previous study applied lesion_GNB, a stroke segmentation classifier, to the resection cavity segmentation problem in epilepsy patients ( Gau et al, 2020 , Griffis et al, 2016 ). While lesion_GNB demonstrated some utility in segmenting resections (median DSC 0.58), our classifier achieved greater segmentation accuracy.…”

ResectVol: A tool to automatically segment and characterize lacunas in brain images

“…Although many studies have focused on segmenting brain lesions, 10 , 16 to the best of our knowledge, not as many have specifically targeted brain lacunas. Gau and colleagues 11 claim to be the first group to have done so. They compared a fully automatic and a semi‐automatic approach against the manual segmentation masks and found a median DSC of 0.58 and 0.78, respectively.…”

“…We randomly selected 60 individuals with temporal lobe epilepsy (TLE) who met inclusion criteria, which is approximately two times the number used in a similar study. 11 After reviewing the quality of MR images, we excluded 4 cases due to MRI artifacts and 2 cases due to MRI signs of gliosis secondary to postoperative infection. In order to avoid the interference of blood and edema in the segmentation, we excluded patients whose postoperative MRI was performed less than five months after surgery (n = 3).…”