Clinical Evaluation of a Multiparametric Deep Learning Model for Glioblastoma Segmentation Using Heterogeneous Magnetic Resonance Imaging Data From Clinical Routine

Perkuhn, Michael; Stavrinou, Pantelis; Thiele, Frank; Shakirin, Georgy; Mohan, Manoj; Garmpis, Dionysios; Kabbasch, Christoph; Borggrefe, Jan

doi:10.1097/rli.0000000000000484

Cited by 58 publications

(72 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Few studies have compared accuracy of automated and manual segmentations of tumour subregions such as necrosis [15]. It is important to measure these subregions as they may have prognostic significance.…”

Section: Discussionmentioning

confidence: 99%

“…Following the procedure from BRATS, automated segmentations of the whole-tumour (WT) region were created from T2-weighted sequences and FLAIR sequences whilst the CER region and NC regions were created from T1-weighted images. PTE and NET were manually delineated from the FLAIR region [15].…”

Section: Tumour Segmentationmentioning

confidence: 99%

“…Each pathway samples different resolutions to lower computational cost [14]. DeepMedic has been shown to have robust segmentation accuracy of tumour subregions when tested and trained on MRI images performed at multiple institutions, with different protocols [15].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep learning for glioblastoma segmentation using preoperative magnetic resonance imaging identifies volumetric features associated with survival

2020

View full text Add to dashboard Cite

Background Measurement of volumetric features is challenging in glioblastoma. We investigate whether volumetric features derived from preoperative MRI using a convolutional neural network–assisted segmentation is correlated with survival. Methods Preoperative MRI of 120 patients were scored using Visually Accessible Rembrandt Images (VASARI) features. We trained and tested a multilayer, multi-scale convolutional neural network on multimodal brain tumour segmentation challenge (BRATS) data, prior to testing on our dataset. The automated labels were manually edited to generate ground truth segmentations. Network performance for our data and BRATS data was compared. Multivariable Cox regression analysis corrected for multiple testing using the false discovery rate was performed to correlate clinical and imaging variables with overall survival. Results Median Dice coefficients in our sample were (1) whole tumour 0.94 (IQR, 0.82–0.98) compared to 0.91 (IQR, 0.83–0.94 p = 0.012), (2) FLAIR region 0.84 (IQR, 0.63–0.95) compared to 0.81 (IQR, 0.69–0.8 p = 0.170), (3) contrast-enhancing region 0.91 (IQR, 0.74–0.98) compared to 0.83 (IQR, 0.78–0.89 p = 0.003) and (4) necrosis region were 0.82 (IQR, 0.47–0.97) compared to 0.67 (IQR, 0.42–0.81 p = 0.005). Contrast-enhancing region/tumour core ratio (HR 4.73 [95% CI, 1.67–13.40], corrected p = 0.017) and necrotic core/tumour core ratio (HR 8.13 [95% CI, 2.06–32.12], corrected p = 0.011) were independently associated with overall survival. Conclusion Semi-automated segmentation of glioblastoma using a convolutional neural network trained on independent data is robust when applied to routine clinical data. The segmented volumes have prognostic significance.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Tumour Segmentationmentioning

confidence: 99%

See 1 more Smart Citation

Deep learning for glioblastoma segmentation using preoperative magnetic resonance imaging identifies volumetric features associated with survival

2020

View full text Add to dashboard Cite

show abstract

“…To determine an "optimal" threshold, two methods were applied. (I) the Youden Index (sensitivity + specificity -1), and (II) the Dice's coefficient 26,27 .…”

Section: Discussionmentioning

confidence: 99%

Comparison of Accuracy of Arrival-Time-Insensitive and Arrival-Time-Sensitive CTP Algorithms for Prediction of Infarct Tissue Volumes

Pennig¹,

Thiele²,

Goertz³

et al. 2020

Sci Rep

Self Cite

View full text Add to dashboard Cite

the purpose of this study was to compare the performance of arrival-time-insensitive (Ati) and arrivaltime-sensitive (AtS) computed tomography perfusion (ctp) algorithms in philips intelliSpace portal (v9, ISP) and to investigate optimal thresholds for ATI regarding the prediction of final infarct volume (FIV). Retrospective, single-center study with 54 patients (mean 67.0 ± 13.1 years, 68.5% male) who received Stroke-CT/CTP-imaging between 2010 and 2018 with occlusion of the middle cerebral artery in the M1-/proximal M2-segment or terminal internal carotid artery. FIV was determined on shortterm follow-up imaging in two patient groups: A) not attempted or failed mechanical thrombectomy (Mt) and B) successful Mt. AtS (default settings) and Ati (full-range of threshold settings regarding FIV prediction) maps were coregistered in 3D with FIV using voxel-wise overlap measurement. Based on an average imaging follow-up of 2.6 ± 2.1 days, the estimation regarding penumbra (group A, ATI: r = 0.63/0.69, ATS: r = 0.64) and infarct core (group B, ATI: r = 0.60/0.68, ATS: r = 0.63) was slightly higher in ATI but the effect was not significant (p > 0.05). Regarding ATI, Tmax (AUC 0.9) was the best estimator of the penumbra (group A), CBF relative to the contralateral hemisphere (AUC 0.80) showed the best estimation of the infarct core (group B). there was a broad range of thresholds of optimal Ati settings in both groups. prediction of fiV with Ati was slightly better compared to AtS. However, this difference was not significant. Since ATI showed a broad range of optimal thresholds, exact thresholds regarding the Ati algorithm should be evaluated in further prospective, clinical studies. Computed tomography perfusion (CTP) represents a valuable adjunct to unenhanced CT and CT-angiography (CTA) in the diagnosis of acute ischemic stroke (AIS), playing a vital role in clinical decision making 1-3. By generating a four-dimensional dataset of the brain, tissue perfusion parameters derived from CTP allow to differentiate between ischemic core and penumbra using selected combinations and thresholds 4. If CTP shows a prevalent mismatch between ischemic core and penumbra, patients can benefit from mechanical thrombectomy (MT) up to 24 hours after AIS onset 1. Moreover, an ischemic core volume of 70 ml is regarded as the critical threshold above which patients show a significantly worse clinical outcome; therefore, accurate core and penumbra measurements derived from CTP are essential 5,6. For many years, the arrival-time-sensitive (ATS) post-processing algorithm represented the standard for CTP, which was thoroughly tested during the last two decades 4,7-9. A potential limitation of ATS is its sensitivity to contrast agent arrival delay and dispersion 7,10 , which can be problematic in patients with reduced cardiac output,

show abstract

“…This is partly because feature extraction involves multiple algorithms and multiple methodologies. For example, Perkuhn et al [22] used 5 3 kernels for feature extraction in four convolutional layers. It would be difficult to summarise such extensive and numerous convolution methodology.…”

Section: Literature Findingsmentioning

confidence: 99%

Convolutional neural networks for brain tumour segmentation

2020

View full text Add to dashboard Cite

The introduction of quantitative image analysis has given rise to fields such as radiomics which have been used to predict clinical sequelae. One growing area of interest for analysis is brain tumours, in particular glioblastoma multiforme (GBM). Tumour segmentation is an important step in the pipeline in the analysis of this pathology. Manual segmentation is often inconsistent as it varies between observers. Automated segmentation has been proposed to combat this issue. Methodologies such as convolutional neural networks (CNNs) which are machine learning pipelines modelled on the biological process of neurons (called nodes) and synapses (connections) have been of interest in the literature. We investigate the role of CNNs to segment brain tumours by firstly taking an educational look at CNNs and perform a literature search to determine an example pipeline for segmentation. We then investigate the future use of CNNs by exploring a novel field-radiomics. This examines quantitative features of brain tumours such as shape, texture, and signal intensity to predict clinical outcomes such as survival and response to therapy.

show abstract

Clinical Evaluation of a Multiparametric Deep Learning Model for Glioblastoma Segmentation Using Heterogeneous Magnetic Resonance Imaging Data From Clinical Routine

Cited by 58 publications

References 28 publications

Deep learning for glioblastoma segmentation using preoperative magnetic resonance imaging identifies volumetric features associated with survival

Deep learning for glioblastoma segmentation using preoperative magnetic resonance imaging identifies volumetric features associated with survival

Comparison of Accuracy of Arrival-Time-Insensitive and Arrival-Time-Sensitive CTP Algorithms for Prediction of Infarct Tissue Volumes

Convolutional neural networks for brain tumour segmentation

Contact Info

Product

Resources

About