Designing image segmentation studies: Statistical power, sample size and reference standard quality

Gibson, Eli; Hu, Yipeng; Huisman, Henkjan; Barratt, Dean C.

doi:10.1016/j.media.2017.07.004

Cited by 19 publications

(16 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Moreover, we believe that the results from this work may corroborate the findings of a number of previous studies in which caution has been raised over the interpretation of the value of some segmentation metrics, and the resulting league table positions in segmentation challenges ( Gibson et al., 2017a , Reinke et al., 2018 ). With evidence from the prostate segmentation in MRIs, we found that a statistically significant difference in the DSC between segmentations produced by two CNNs, does not necessarily lead to any detectable impact in other computational tasks within a clinical workflow that use these segmentations.…”

Section: Discussionsupporting

confidence: 85%

“…Partly limited by the test data size of 30 images provided in the PROMISE12 Challenge, the diminishing statistically significant differences among top performing segmentation algorithms ( Gibson et al., 2017a ) can complicate interpreting these differences, if any, in segmentation accuracy. In other research fields, however, examples of networks that demonstrated statistical differences between different architectures include the introduction of residual networks ( He et al., 2016 ) and densely connected networks ( Huang et al., 2017 ).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automatic segmentation of prostate MRI using convolutional neural networks: Investigating the impact of network architecture on the accuracy of volume measurement and MRI-ultrasound registration

Ghavami

Gibson

et al. 2019

Medical Image Analysis

Self Cite

View full text Add to dashboard Cite

Highlights Six openly-available neural networks for prostate MR segmentation were compared, based on experiment results using 232-patient clinical imaging and label data. Statistically significant difference in segmentation accuracy was found in one network. No statistically significant difference was found in subsequent clinical applications, for volume estimation or image registration where these segmentations were used.

show abstract

Section: Discussionsupporting

confidence: 85%

Section: Introductionmentioning

confidence: 99%

Automatic segmentation of prostate MRI using convolutional neural networks: Investigating the impact of network architecture on the accuracy of volume measurement and MRI-ultrasound registration

Ghavami

Gibson

et al. 2019

Medical Image Analysis

Self Cite

View full text Add to dashboard Cite

show abstract

“…Furthermore the resection cavity should be clearly present on visual inspection. A total of 30 patients were retrospectively included in the study, which is in line with recommendations for the evaluation of segmentation accuracy from Gibson et al [25]. All patients received adjuvant concomitant chemo-radiotherapy with Temozolomide.…”

Section: Patientsmentioning

confidence: 99%

Fully automated brain resection cavity delineation for radiation target volume definition in glioblastoma patients using deep learning

et al. 2020

View full text Add to dashboard Cite

Background: Automated brain tumor segmentation methods are computational algorithms that yield tumor delineation from, in this case, multimodal magnetic resonance imaging (MRI). We present an automated segmentation method and its results for resection cavity (RC) in glioblastoma multiforme (GBM) patients using deep learning (DL) technologies. Methods: Post-operative, T1w with and without contrast, T2w and fluid attenuated inversion recovery MRI studies of 30 GBM patients were included. Three radiation oncologists manually delineated the RC to obtain a reference segmentation. We developed a DL cavity segmentation method, which utilizes all four MRI sequences and the reference segmentation to learn to perform RC delineations. We evaluated the segmentation method in terms of Dice coefficient (DC) and estimated volume measurements. Results: Median DC of the three radiation oncologist were 0.85 (interquartile range [IQR]: 0.08), 0.84 (IQR: 0.07), and 0.86 (IQR: 0.07). The results of the automatic segmentation compared to the three different raters were 0.83 (IQR: 0.14), 0.81 (IQR: 0.12), and 0.81 (IQR: 0.13) which was significantly lower compared to the DC among raters (chisquare = 11.63, p = 0.04). We did not detect a statistically significant difference of the measured RC volumes for the different raters and the automated method (Kruskal-Wallis test: chi-square = 1.46, p = 0.69). The main sources of error were due to signal inhomogeneity and similar intensity patterns between cavity and brain tissues. Conclusions: The proposed DL approach yields promising results for automated RC segmentation in this proof of concept study. Compared to human experts, the DC are still subpar.

show abstract

“…LeukNet was designed after analyzing the previously described results, where VGG-16 and VGG-19 architectures presented the best outcomes, with similar values for the mDFT approach in the ALL-IDB2 dataset. Therefore, we performed the Student's t-test [37] to statistically compare the results at a significance level of 5%. From the test performed, we found that the results were equivalent.…”

Section: A Models and Fine-tuning Evaluationmentioning

confidence: 99%

"LeukNet" - A Model of Convolutional Neural Network for the Diagnosis of Leukemia

Vogado¹,

Veras²,

Aires³

2020

Anais Estendidos Da Conference on Graphics, Patterns and Images (SIBRAPI Estendido 2020)

View full text Add to dashboard Cite

Leukemia is a disorder that affects the bone marrow, causing uncontrolled production of leukocytes, impairing the transport of oxygen and causing blood coagulation problems. In this article, we propose a new computational tool, named LeukNet, a Convolutional Neural Network (CNN) architecture based on the VGG-16 convolutional blocks, to facilitate the leukemia diagnosis from blood smear images. We evaluated different architectures and fine-tuning methods using 18 datasets containing 3536 images with distinct characteristics of color, texture, contrast, and resolution. Additionally, data augmentation operations were applied to increase the training set by up to 20 times. The k-fold cross-validation (k = 5) results achieved 98.28% of accuracy. A cross-dataset validation technique, named LeaveOne-Dataset-Out Cross-Validation (LODOCV), is also proposed to evaluate the developed model’s generalization capability. The accuracy of using LODOCV on the ALL-IDB 1, ALL-IDB 2, and UFG datasets was 97.04%, 82.46%, and 70.24%, respectively, overcoming the current state-of-the-art results and offering new guidelines for image-based computer-aided diagnosis (CAD) systems in this area.

show abstract

Designing image segmentation studies: Statistical power, sample size and reference standard quality

Cited by 19 publications

References 33 publications

Automatic segmentation of prostate MRI using convolutional neural networks: Investigating the impact of network architecture on the accuracy of volume measurement and MRI-ultrasound registration

Automatic segmentation of prostate MRI using convolutional neural networks: Investigating the impact of network architecture on the accuracy of volume measurement and MRI-ultrasound registration

Fully automated brain resection cavity delineation for radiation target volume definition in glioblastoma patients using deep learning

"LeukNet" - A Model of Convolutional Neural Network for the Diagnosis of Leukemia

Contact Info

Product

Resources

About