Sabine Heiland
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sabine Heiland has a total of 86 co-authors. Here are the top co-authors by number of shared publications:
- Martin Bendszus
Abteilung für Neuroradiologie, Universitätsklinikum Heidelberg, Heidelberg, Germany
14 shared publications - Tim Godel
Department of Neuroradiology, Heidelberg University Hospital, Im Neuenheimer Feld 400, 69120 Heidelberg, Germany
5 shared publications - Tim Hilgenfeld
Department of Neuroradiology, University Hospital Heidelberg, Heidelberg, Germany
5 shared publications - Amir Abdollahi
Clinical Cooperation Unit Translational Radiation Oncology, National Center for Tumor Diseases (NCT), Heidelberg University Hospital (UKHD) and German Cancer Research Center (DKFZ), Heidelberg, Germany
4 shared publications - Andreas Kudak
Universitätsklinikum Heidelberg / HIRO
4 shared publications - Christel Herold-Mende
Department of Neurosurgery, Heidelberg University Hospital, Heidelberg, Germany
4 shared publications
Latest Publications
- 13 October 2025
- Published by Wolters Kluwer Health in Investigative Radiology
- Vol. 61 (6) , 387-394
Improving Spatial Accuracy of MRI in the Presence of Metal-induced Geometric Distortions
Objectives: To identify sequences and protocols for minimal metal-induced geometric distortion (MD) for improved spatial accuracy in MRI. Materials and Methods: A 3D lattice phantom containing a stainless-steel bracket or a crown-supported titanium implant was scanned using 6 MRI sequences (TSE, SEMAC, CS-SEMAC, SPACE, VIBE, and research sequence MSVAT-SPACE) in a 3T system. MD was assessed at 9360 crossing points as Euclidean distance using a standardized algorithm. MD was analyzed by total MD, affected volume (AV) at various thresholds, and directional dependency. Statistical analysis was performed by one-way ANOVA. Results: For the stainless-steel bracket, TSE showed the highest total MD among all sequences (2187±297 mm, P4 mm), with 467 mL AV at MD>0.5 mm. CS-SEMAC and SEMAC yielded the lowest MD among all sequences (469±75 mm and 502±154 mm, P0.5 mm. 3D sequences exhibited intermediate performance with no significant difference (MSVAT-SPACE/VIBE/SPACE: 1569±204 mm/1137±71 mm/1513±143 mm; P>0.08). For the crown-supported titanium implant, all sequences showed reduced MD (PP=0.0071), while distortions in phase direction were consistently lower. Conclusions: Susceptibility artifact reduction sequences, particularly SEMAC and CS-SEMAC, effectively reduce total-MD by 79% and AV by 90%. Distortion varies by encoding direction and is most severe along the frequency-encoding axis, highlighting the importance of sequence and parameter selection for accurate MRI near metal implants.
- 13 October 2025
- Published by Wolters Kluwer Health in Investigative Radiology
- Vol. 61 (6) , 387-394
Improving Spatial Accuracy of MRI in the Presence of Metal-induced Geometric Distortions
Objectives: To identify sequences and protocols for minimal metal-induced geometric distortion (MD) for improved spatial accuracy in MRI. Materials and Methods: A 3D lattice phantom containing a stainless-steel bracket or a crown-supported titanium implant was scanned using 6 MRI sequences (TSE, SEMAC, CS-SEMAC, SPACE, VIBE, and research sequence MSVAT-SPACE) in a 3T system. MD was assessed at 9360 crossing points as Euclidean distance using a standardized algorithm. MD was analyzed by total MD, affected volume (AV) at various thresholds, and directional dependency. Statistical analysis was performed by one-way ANOVA. Results: For the stainless-steel bracket, TSE showed the highest total MD among all sequences (2187±297 mm, P4 mm), with 467 mL AV at MD>0.5 mm. CS-SEMAC and SEMAC yielded the lowest MD among all sequences (469±75 mm and 502±154 mm, P0.5 mm. 3D sequences exhibited intermediate performance with no significant difference (MSVAT-SPACE/VIBE/SPACE: 1569±204 mm/1137±71 mm/1513±143 mm; P>0.08). For the crown-supported titanium implant, all sequences showed reduced MD (PP=0.0071), while distortions in phase direction were consistently lower. Conclusions: Susceptibility artifact reduction sequences, particularly SEMAC and CS-SEMAC, effectively reduce total-MD by 79% and AV by 90%. Distortion varies by encoding direction and is most severe along the frequency-encoding axis, highlighting the importance of sequence and parameter selection for accurate MRI near metal implants.
- Open Access
- 25 March 2023
- Published by MDPI AG in Diagnostics
- Vol. 13 (7) , 1237
Quantitative MR Neurography in Multifocal Motor Neuropathy and Amyotrophic Lateral Sclerosis
Background: The aim of this study was to assess the phenotype of multifocal motor neuropathy (MMN) and amyotrophic lateral sclerosis (ALS) in quantitative MR neurography. Methods: In this prospective study, 22 patients with ALS, 8 patients with MMN, and 10 healthy volunteers were examined with 3T MR neurography, using a high-resolution fat-saturated T2-weighted sequence, diffusion-tensor imaging (DTI), and a multi-echo T2-relaxometry sequence. The quantitative biomarkers fractional anisotropy (FA), radial and axial diffusivity (RD, AD), mean diffusivity (MD), cross-sectional area (CSA), T2-relaxation time, and proton spin density (PSD) were measured in the tibial nerve at the thigh and calf, and in the median, radial, and ulnar nerves at the mid-upper arm. Results: MMN showed a characteristic imaging pattern of decreased FA (p = 0.018), increased RD (p = 0.014), increased CSA (p < 0.001), increased T2-relaxation time (p < 0.001), and increased PSD (p = 0.025) in the upper arm nerves compared to ALS and controls. ALS patients did not differ from controls in any imaging marker, nor were there any group differences in the tibial nerve (p > 0.05). Conclusions: MMN shows a characteristic pattern of quantitative DTI and T2-relaxometry parameters in the upper-arm nerves, primarily indicating demyelination. Peripheral nerve changes in ALS seem to be below the detection level of current state-of-the-art quantitative MR neurography.
- 25 March 2023
- Published by MDPI AG in Diagnostics
- Vol. 13 (7) , 1237
Open AccessQuantitative MR Neurography in Multifocal Motor Neuropathy and Amyotrophic Lateral Sclerosis
Background: The aim of this study was to assess the phenotype of multifocal motor neuropathy (MMN) and amyotrophic lateral sclerosis (ALS) in quantitative MR neurography. Methods: In this prospective study, 22 patients with ALS, 8 patients with MMN, and 10 healthy volunteers were examined with 3T MR neurography, using a high-resolution fat-saturated T2-weighted sequence, diffusion-tensor imaging (DTI), and a multi-echo T2-relaxometry sequence. The quantitative biomarkers fractional anisotropy (FA), radial and axial diffusivity (RD, AD), mean diffusivity (MD), cross-sectional area (CSA), T2-relaxation time, and proton spin density (PSD) were measured in the tibial nerve at the thigh and calf, and in the median, radial, and ulnar nerves at the mid-upper arm. Results: MMN showed a characteristic imaging pattern of decreased FA (p = 0.018), increased RD (p = 0.014), increased CSA (p < 0.001), increased T2-relaxation time (p < 0.001), and increased PSD (p = 0.025) in the upper arm nerves compared to ALS and controls. ALS patients did not differ from controls in any imaging marker, nor were there any group differences in the tibial nerve (p > 0.05). Conclusions: MMN shows a characteristic pattern of quantitative DTI and T2-relaxometry parameters in the upper-arm nerves, primarily indicating demyelination. Peripheral nerve changes in ALS seem to be below the detection level of current state-of-the-art quantitative MR neurography.
- Open Access
MR-Class: A Python Tool for Brain MR Image Classification Utilizing One-vs-All DCNNs to Deal with the Open-Set Recognition Problem
Background: MR image classification in datasets collected from multiple sources is complicated by inconsistent and missing DICOM metadata. Therefore, we aimed to establish a method for the efficient automatic classification of MR brain sequences. Methods: Deep convolutional neural networks (DCNN) were trained as one-vs-all classifiers to differentiate between six classes: T1 weighted (w), contrast-enhanced T1w, T2w, T2w-FLAIR, ADC, and SWI. Each classifier yields a probability, allowing threshold-based and relative probability assignment while excluding images with low probability (label: unknown, open-set recognition problem). Data from three high-grade glioma (HGG) cohorts was assessed; C1 (320 patients, 20,101 MRI images) was used for training, while C2 (197, 11,333) and C3 (256, 3522) were for testing. Two raters manually checked images through an interactive labeling tool. Finally, MR-Class’ added value was evaluated via radiomics model performance for progression-free survival (PFS) prediction in C2, utilizing the concordance index (C-I). Results: Approximately 10% of annotation errors were observed in each cohort between the DICOM series descriptions and the derived labels. MR-Class accuracy was 96.7% [95% Cl: 95.8, 97.3] for C2 and 94.4% [93.6, 96.1] for C3. A total of 620 images were misclassified; manual assessment of those frequently showed motion artifacts or alterations of anatomy by large tumors. Implementation of MR-Class increased the PFS model C-I by 14.6% on average, compared to a model trained without MR-Class. Conclusions: We provide a DCNN-based method for the sequence classification of brain MR images and demonstrate its usability in two independent HGG datasets.
Open AccessMR-Class: A Python Tool for Brain MR Image Classification Utilizing One-vs-All DCNNs to Deal with the Open-Set Recognition Problem
Background: MR image classification in datasets collected from multiple sources is complicated by inconsistent and missing DICOM metadata. Therefore, we aimed to establish a method for the efficient automatic classification of MR brain sequences. Methods: Deep convolutional neural networks (DCNN) were trained as one-vs-all classifiers to differentiate between six classes: T1 weighted (w), contrast-enhanced T1w, T2w, T2w-FLAIR, ADC, and SWI. Each classifier yields a probability, allowing threshold-based and relative probability assignment while excluding images with low probability (label: unknown, open-set recognition problem). Data from three high-grade glioma (HGG) cohorts was assessed; C1 (320 patients, 20,101 MRI images) was used for training, while C2 (197, 11,333) and C3 (256, 3522) were for testing. Two raters manually checked images through an interactive labeling tool. Finally, MR-Class’ added value was evaluated via radiomics model performance for progression-free survival (PFS) prediction in C2, utilizing the concordance index (C-I). Results: Approximately 10% of annotation errors were observed in each cohort between the DICOM series descriptions and the derived labels. MR-Class accuracy was 96.7% [95% Cl: 95.8, 97.3] for C2 and 94.4% [93.6, 96.1] for C3. A total of 620 images were misclassified; manual assessment of those frequently showed motion artifacts or alterations of anatomy by large tumors. Implementation of MR-Class increased the PFS model C-I by 14.6% on average, compared to a model trained without MR-Class. Conclusions: We provide a DCNN-based method for the sequence classification of brain MR images and demonstrate its usability in two independent HGG datasets.
- Open Access
MR Intensity Normalization Methods Impact Sequence Specific Radiomics Prognostic Model Performance in Primary and Recurrent High-Grade Glioma
Purpose: This study investigates the impact of different intensity normalization (IN) methods on the overall survival (OS) radiomics models’ performance of MR sequences in primary (pHGG) and recurrent high-grade glioma (rHGG). Methods: MR scans acquired before radiotherapy were retrieved from two independent cohorts (rHGG C1: 197, pHGG C2: 141) from multiple scanners (15, 14). The sequences are T1 weighted (w), contrast-enhanced T1w (T1wce), T2w, and T2w-FLAIR. Sequence-specific significant features (SF) associated with OS, extracted from the tumour volume, were derived after applying 15 different IN methods. Survival analyses were conducted using Cox proportional hazard (CPH) and Poisson regression (POI) models. A ranking score was assigned based on the 10-fold cross-validated (CV) concordance index (C-I), mean square error (MSE), and the Akaike information criterion (AICs), to evaluate the methods’ performance. Results: Scatter plots of the 10-CV C-I and MSE against the AIC showed an impact on the survival predictions between the IN methods and MR sequences (C1/C2 C-I range: 0.62–0.71/0.61–0.72, MSE range: 0.20–0.42/0.13–0.22). White stripe showed stable results for T1wce (C1/C2 C-I: 0.71/0.65, MSE: 0.21/0.14). Combat (0.68/0.62, 0.22/0.15) and histogram matching (HM, 0.67/0.64, 0.22/0.15) showed consistent prediction results for T2w models. They were also the top-performing methods for T1w in C2 (Combat: 0.67, 0.13; HM: 0.67, 0.13); however, only HM achieved high predictions in C1 (0.66, 0.22). After eliminating IN impacted SF using Spearman’s rank-order correlation coefficient, a mean decrease in the C-I and MSE of 0.05 and 0.03 was observed in all four sequences. Conclusion: The IN method impacted the predictive power of survival models; thus, performance is sequence-dependent.
Open AccessMR Intensity Normalization Methods Impact Sequence Specific Radiomics Prognostic Model Performance in Primary and Recurrent High-Grade Glioma
Purpose: This study investigates the impact of different intensity normalization (IN) methods on the overall survival (OS) radiomics models’ performance of MR sequences in primary (pHGG) and recurrent high-grade glioma (rHGG). Methods: MR scans acquired before radiotherapy were retrieved from two independent cohorts (rHGG C1: 197, pHGG C2: 141) from multiple scanners (15, 14). The sequences are T1 weighted (w), contrast-enhanced T1w (T1wce), T2w, and T2w-FLAIR. Sequence-specific significant features (SF) associated with OS, extracted from the tumour volume, were derived after applying 15 different IN methods. Survival analyses were conducted using Cox proportional hazard (CPH) and Poisson regression (POI) models. A ranking score was assigned based on the 10-fold cross-validated (CV) concordance index (C-I), mean square error (MSE), and the Akaike information criterion (AICs), to evaluate the methods’ performance. Results: Scatter plots of the 10-CV C-I and MSE against the AIC showed an impact on the survival predictions between the IN methods and MR sequences (C1/C2 C-I range: 0.62–0.71/0.61–0.72, MSE range: 0.20–0.42/0.13–0.22). White stripe showed stable results for T1wce (C1/C2 C-I: 0.71/0.65, MSE: 0.21/0.14). Combat (0.68/0.62, 0.22/0.15) and histogram matching (HM, 0.67/0.64, 0.22/0.15) showed consistent prediction results for T2w models. They were also the top-performing methods for T1w in C2 (Combat: 0.67, 0.13; HM: 0.67, 0.13); however, only HM achieved high predictions in C1 (0.66, 0.22). After eliminating IN impacted SF using Spearman’s rank-order correlation coefficient, a mean decrease in the C-I and MSE of 0.05 and 0.03 was observed in all four sequences. Conclusion: The IN method impacted the predictive power of survival models; thus, performance is sequence-dependent.
- Open Access
- 26 January 2023
- Published by Springer Nature in Journal of Orofacial Orthopedics / Fortschritte der Kieferorthopädie
- Vol. 85 (4) , 279-288
In vivo assessment of artefacts in MRI images caused by conventional twistflex and various fixed orthodontic CAD/CAM retainers
Purpose: To assess magnetic resonance imaging (MRI) artefacts caused by different computer-aided design/computer-aided manufacturing (CAD/CAM) retainers in comparison with conventional hand bent stainless steel twistflex retainers in vivo. Materials and methods: MRI scans (3 Tesla) were performed on a male volunteer with different CAD/CAM retainers (cobalt–chromium, CoCr; nickel–titanium, NiTi; grade 5 titanium, Ti5) and twistflex retainers inserted. A total of 126 landmarks inside and outside the retainer area (RA; from canine to canine) were evaluated by two blinded radiologists using an established five-point visibility scoring (1: excellent, 2: good, 3: moderate, 4: poor, 5: not visible). Friedman and two-tailed Wilcoxon tests were used for statistical analysis (significance level: p < 0.05). Results: Twistflex retainers had the strongest impact on the visibility of all landmarks inside (4.0 ± 1.5) and outside the RA (1.7 ± 1.2). In contrast, artefacts caused by CAD/CAM retainers were limited to the dental area inside the RA (CoCr: 2.2 ± 1.2) or did not impair MRI-based diagnostics in a clinically relevant way (NiTi: 1.0 ± 0.1; Ti5: 1.4 ± 0.6). Conclusion: The present study on a single test person demonstrates that conventional stainless steel twistflex retainers can severely impair the diagnostic value in head/neck and dental MRI. By contrast, CoCr CAD/CAM retainers can cause artefacts which only slightly impair dental MRI but not head/neck MRI, whereas NiTi and Ti5 CAD/CAM might be fully compatible with both head/neck and dental MRI.
- 26 January 2023
- Published by Springer Nature in Journal of Orofacial Orthopedics / Fortschritte der Kieferorthopädie
- Vol. 85 (4) , 279-288
Open AccessIn vivo assessment of artefacts in MRI images caused by conventional twistflex and various fixed orthodontic CAD/CAM retainers
Purpose: To assess magnetic resonance imaging (MRI) artefacts caused by different computer-aided design/computer-aided manufacturing (CAD/CAM) retainers in comparison with conventional hand bent stainless steel twistflex retainers in vivo. Materials and methods: MRI scans (3 Tesla) were performed on a male volunteer with different CAD/CAM retainers (cobalt–chromium, CoCr; nickel–titanium, NiTi; grade 5 titanium, Ti5) and twistflex retainers inserted. A total of 126 landmarks inside and outside the retainer area (RA; from canine to canine) were evaluated by two blinded radiologists using an established five-point visibility scoring (1: excellent, 2: good, 3: moderate, 4: poor, 5: not visible). Friedman and two-tailed Wilcoxon tests were used for statistical analysis (significance level: p < 0.05). Results: Twistflex retainers had the strongest impact on the visibility of all landmarks inside (4.0 ± 1.5) and outside the RA (1.7 ± 1.2). In contrast, artefacts caused by CAD/CAM retainers were limited to the dental area inside the RA (CoCr: 2.2 ± 1.2) or did not impair MRI-based diagnostics in a clinically relevant way (NiTi: 1.0 ± 0.1; Ti5: 1.4 ± 0.6). Conclusion: The present study on a single test person demonstrates that conventional stainless steel twistflex retainers can severely impair the diagnostic value in head/neck and dental MRI. By contrast, CoCr CAD/CAM retainers can cause artefacts which only slightly impair dental MRI but not head/neck MRI, whereas NiTi and Ti5 CAD/CAM might be fully compatible with both head/neck and dental MRI.
- Open Access
- 10 January 2023
- Published by MDPI AG in Preprints.org
MR Intensity Normalization Methods Impact Sequence Specific Radiomics Prognostic Model Performance in Primary and Recurrent High-Grade Glioma
Purpose: This study investigates the impact of different intensity normalization (IN) methods on the overall survival (OS) radiomics models’ performance of MR sequences in primary (pHGG) and recurrent high-grade glioma (rHGG). Methods: MR scans acquired before radiotherapy were retrieved from two independent cohorts (rHGG C1: 197, pHGG C2: 141) from multiple scanners (15, 14). The sequences are T1 weighted (w), contrast-enhanced T1w (T1wce), T2w, and T2w-FLAIR. Sequence-specific significant features (SF) associated with OS, extracted from the tumour volume, were derived after applying 15 different IN methods. Survival analyses were conducted using Cox proportional hazard (CPH) and Poisson regression (POI) models. A ranking score was assigned based on the 10-fold cross-validated (CV) concordance index (C-I), mean square error (MSE), and the Akaike information criterion (AICs), to evaluate the methods’ performance. Results: Scatter plots of the 10-CV C-I and MSE against the AIC showed an impact on the survival predictions between the IN methods and MR sequences (C1/C2 C-I range: 0.62-0.71/0.61-0.72, MSE range: 0.20-0.42/0.13-0.22). White stripe showed stable results for T1wce (C1/C2 C-I: 0.71/0.65, MSE: 0.21/0.14). Combat (0.68/0.62, 0.22/0.15) and histogram matching (HM, 0.67/0.64, 0.22/0.15) showed consistent prediction results for T2w models. They were also the top-performing methods for T1w in C2 (Combat: 0.67, 0.13; HM: 0.67, 0.13), however, only HM achieved high predictions in C1 (0.66, 0.22). After eliminating IN impacted SF using Spearman’s rank-order correlation coefficient, a mean decrease in the C-I and MSE of 0.05 and 0.03 was observed in all four sequences. Conclusion: The IN method impacted the predictive power of survival models. Thus, performance is sequence-dependent.
- 10 January 2023
- Published by MDPI AG in Preprints.org
Open AccessMR Intensity Normalization Methods Impact Sequence Specific Radiomics Prognostic Model Performance in Primary and Recurrent High-Grade Glioma
Purpose: This study investigates the impact of different intensity normalization (IN) methods on the overall survival (OS) radiomics models’ performance of MR sequences in primary (pHGG) and recurrent high-grade glioma (rHGG). Methods: MR scans acquired before radiotherapy were retrieved from two independent cohorts (rHGG C1: 197, pHGG C2: 141) from multiple scanners (15, 14). The sequences are T1 weighted (w), contrast-enhanced T1w (T1wce), T2w, and T2w-FLAIR. Sequence-specific significant features (SF) associated with OS, extracted from the tumour volume, were derived after applying 15 different IN methods. Survival analyses were conducted using Cox proportional hazard (CPH) and Poisson regression (POI) models. A ranking score was assigned based on the 10-fold cross-validated (CV) concordance index (C-I), mean square error (MSE), and the Akaike information criterion (AICs), to evaluate the methods’ performance. Results: Scatter plots of the 10-CV C-I and MSE against the AIC showed an impact on the survival predictions between the IN methods and MR sequences (C1/C2 C-I range: 0.62-0.71/0.61-0.72, MSE range: 0.20-0.42/0.13-0.22). White stripe showed stable results for T1wce (C1/C2 C-I: 0.71/0.65, MSE: 0.21/0.14). Combat (0.68/0.62, 0.22/0.15) and histogram matching (HM, 0.67/0.64, 0.22/0.15) showed consistent prediction results for T2w models. They were also the top-performing methods for T1w in C2 (Combat: 0.67, 0.13; HM: 0.67, 0.13), however, only HM achieved high predictions in C1 (0.66, 0.22). After eliminating IN impacted SF using Spearman’s rank-order correlation coefficient, a mean decrease in the C-I and MSE of 0.05 and 0.03 was observed in all four sequences. Conclusion: The IN method impacted the predictive power of survival models. Thus, performance is sequence-dependent.
- Open Access
- 9 January 2023
- Published by MDPI AG in Preprints.org
MR-Class: A Python Tool for Brain Mr Image Classification Utilizing One-Vs-All DCNNS to Deal With the Open-Set Recognition Problem
Background: MR image classification in datasets collected from multiple sources is complicated by inconsistent and missing DICOM metadata. Therefore, we aimed to establish a method for the efficient automatic classification of MR brain sequences. Methods: Deep convolutional neural networks (DCNN) were trained as one-vs-all classifiers to differentiate between six classes, T1 weighted (w), contrast-enhanced T1w, T2w, T2w-FLAIR, ADC, and SWI. Each classifier yields a probability, allowing threshold-based and relative probability assignment while excluding images with low probability (label: unknown, open-set recognition problem). Data from three high-grade glioma (HGG) cohorts was assessed; C1 (320 patients, 20101 MRI images) was used for training, while C2 (197, 11333) and C3 (256, 3522) were for testing. Two raters manually checked images through an interactive labeling tool. Finally, MR-Class' added value was evaluated via radiomics models' performance for progression-free survival (PFS) prediction in C2, utilizing the concordance index (C-I). Results: Approximately 10% of annotation errors were observed in each cohort between the DICOM series descriptions and the derived labels. MR-Class accuracy was 96.7% [95%-Cl: 95.8, 97.3] for C2 and 94.4% [93.6, 96.1] for C3. 620 images were misclassified; Manual assessment of those frequently showed motion artifacts or alterations of anatomy by large tumors. Implementation of MR-Class increased on average the PFS model C-I by 14.6% compared to a model trained without MR-Class. Conclusions: We provide a DCNN-based method for sequence classification of brain MR images and demonstrate its usability in two independent HGG datasets.
- 9 January 2023
- Published by MDPI AG in Preprints.org
Open AccessMR-Class: A Python Tool for Brain Mr Image Classification Utilizing One-Vs-All DCNNS to Deal With the Open-Set Recognition Problem
Background: MR image classification in datasets collected from multiple sources is complicated by inconsistent and missing DICOM metadata. Therefore, we aimed to establish a method for the efficient automatic classification of MR brain sequences. Methods: Deep convolutional neural networks (DCNN) were trained as one-vs-all classifiers to differentiate between six classes, T1 weighted (w), contrast-enhanced T1w, T2w, T2w-FLAIR, ADC, and SWI. Each classifier yields a probability, allowing threshold-based and relative probability assignment while excluding images with low probability (label: unknown, open-set recognition problem). Data from three high-grade glioma (HGG) cohorts was assessed; C1 (320 patients, 20101 MRI images) was used for training, while C2 (197, 11333) and C3 (256, 3522) were for testing. Two raters manually checked images through an interactive labeling tool. Finally, MR-Class' added value was evaluated via radiomics models' performance for progression-free survival (PFS) prediction in C2, utilizing the concordance index (C-I). Results: Approximately 10% of annotation errors were observed in each cohort between the DICOM series descriptions and the derived labels. MR-Class accuracy was 96.7% [95%-Cl: 95.8, 97.3] for C2 and 94.4% [93.6, 96.1] for C3. 620 images were misclassified; Manual assessment of those frequently showed motion artifacts or alterations of anatomy by large tumors. Implementation of MR-Class increased on average the PFS model C-I by 14.6% compared to a model trained without MR-Class. Conclusions: We provide a DCNN-based method for sequence classification of brain MR images and demonstrate its usability in two independent HGG datasets.
- 15 August 2022
- Published by Wolters Kluwer Health in Investigative Radiology
- Vol. 58 (2) , 173-179
Peripheral Nerve Involvement at First Diagnosis of Multiple Sclerosis
Objectives The aim of this study was to assess peripheral nerve involvement in patients with multiple sclerosis (MS) at first clinical presentation using quantitative magnetic resonance (MR) neurography in correlation with clinical, laboratory, electrophysiological, and central nervous MR imaging data. Materials and Methods In this prospective monocentric study, 30 patients first diagnosed with MS according to the McDonald criteria (19 women; mean age, 32.4 ± 8.8 years) and 30 age- and sex-matched healthy volunteers were examined with high-resolution 3 T MR neurography using a dual-echo T2-relaxometry sequence covering the tibial and peroneal nerves from proximal thigh to distal calf. Magnetic resonance biomarkers of T2 relaxation time (T2app), proton spin density (PSD), and nerve cross-sectional area (CSA) were correlated with clinical symptoms, intrathecal immunoglobulin (Ig) synthesis, nerve conduction study, and lesion load on brain and spine MR imaging. The diagnostic accuracy of MR biomarkers was assessed using receiver-operating characteristic curves. Results Diffuse nerve changes were detected along the tibial and peroneal nerves in MS patients, who showed decreased PSD (P < 0.001), increased T2app (P < 0.001), and smaller tibial nerve CSA (P < 0.001) compared with healthy subjects. Tibial PSD was identified as best parameter separating patients from controls (area under the curve = 0.876). Intrathecal IgG and IgM synthesis correlated with PSD values (r = −0.44, P = 0.016, and r = −0.42, P = 0.022). Contrast-enhancement of brain or spine lesions was related to larger tibial and peroneal CSA (P < 0.001, P = 0.033). Abnormal electrophysiology correlated with higher tibial and peroneal T2app (P < 0.001 and P = 0.033), lower tibial and peroneal PSD (P = 0.018 and P = 0.002), and smaller peroneal CSA (P < 0.001). Conclusions Quantitative MR neurography reveals peripheral nerve changes in patients with initial diagnosis of MS. Correlation of imaging findings with intrathecal immunoglobulin synthesis may indicate a primary coaffection of the peripheral nervous system in MS.
- 15 August 2022
- Published by Wolters Kluwer Health in Investigative Radiology
- Vol. 58 (2) , 173-179
Peripheral Nerve Involvement at First Diagnosis of Multiple Sclerosis
Objectives The aim of this study was to assess peripheral nerve involvement in patients with multiple sclerosis (MS) at first clinical presentation using quantitative magnetic resonance (MR) neurography in correlation with clinical, laboratory, electrophysiological, and central nervous MR imaging data. Materials and Methods In this prospective monocentric study, 30 patients first diagnosed with MS according to the McDonald criteria (19 women; mean age, 32.4 ± 8.8 years) and 30 age- and sex-matched healthy volunteers were examined with high-resolution 3 T MR neurography using a dual-echo T2-relaxometry sequence covering the tibial and peroneal nerves from proximal thigh to distal calf. Magnetic resonance biomarkers of T2 relaxation time (T2app), proton spin density (PSD), and nerve cross-sectional area (CSA) were correlated with clinical symptoms, intrathecal immunoglobulin (Ig) synthesis, nerve conduction study, and lesion load on brain and spine MR imaging. The diagnostic accuracy of MR biomarkers was assessed using receiver-operating characteristic curves. Results Diffuse nerve changes were detected along the tibial and peroneal nerves in MS patients, who showed decreased PSD (P < 0.001), increased T2app (P < 0.001), and smaller tibial nerve CSA (P < 0.001) compared with healthy subjects. Tibial PSD was identified as best parameter separating patients from controls (area under the curve = 0.876). Intrathecal IgG and IgM synthesis correlated with PSD values (r = −0.44, P = 0.016, and r = −0.42, P = 0.022). Contrast-enhancement of brain or spine lesions was related to larger tibial and peroneal CSA (P < 0.001, P = 0.033). Abnormal electrophysiology correlated with higher tibial and peroneal T2app (P < 0.001 and P = 0.033), lower tibial and peroneal PSD (P = 0.018 and P = 0.002), and smaller peroneal CSA (P < 0.001). Conclusions Quantitative MR neurography reveals peripheral nerve changes in patients with initial diagnosis of MS. Correlation of imaging findings with intrathecal immunoglobulin synthesis may indicate a primary coaffection of the peripheral nervous system in MS.
- Open Access
- 28 June 2022
- Published by MDPI AG in Diagnostics
- Vol. 12 (7) , 1570
Dorsal Root Ganglia Volume—Normative Values, Correlation with Demographic Determinants and Reliability of Three Different Methods of Volumetry
Background: Dorsal root ganglia (DRG) volume assessment by MR-Neurography (MRN) has evolved to an important imaging marker in the diagnostic workup of various peripheral neuropathies and pain syndromes. The aim of this study was (1) to assess normal values of DRG volume and correlations with demographic determinants and (2) to quantify the inter-reader and inter-method reliability of three different methods of DRG volumetry. Methods: Sixty healthy subjects (mean age: 59.1, range 23–79) were examined using a 3D T2-weighted MRN of the lumbosacral plexus at 3 Tesla. Normal values of DRG L3 to S2 were obtained after exact volumetry based on manual 3D segmentation and correlations with demographic variables were assessed. For the assessment of inter-reader and inter-method reliability, DRG volumes in a subset of 25 participants were measured by two independent readers, each applying (1) exact volumetry based on 3D segmentation, (2) axis-corrected, and (3) non-axis-corrected volume estimation. Intraclass correlation coefficients were reported and the Bland–Altman analysis was conducted. Results: Mean DRG volumes ranged from 124.8 mm3 for L3 to 323.3 mm3 for S1 and did not differ between right and left DRG. DRG volume (mean of L3 to S1) correlated with body height (r = 0.42; p = 0.0008) and weight (r = 0.34; p = 0.0087). DRG of men were larger than of women (p = 0.0002); however, no difference remained after correction for body height. Inter-reader reliability was high for all three methods but best for exact volumetry (ICC = 0.99). While axis-corrected estimation was not associated with a relevant bias, non-axis-corrected estimation systematically overestimated DRG volume by on average of 15.55 mm3 (reader 1) or 18.00 mm3 (reader 2) when compared with exact volumetry. Conclusion: The here presented normal values of lumbosacral DRG volume and the correlations with height and weight may be considered in future disease specific studies and possible clinical applications. Exact volumetry was most reliable and should be considered the gold standard. However, the reliability of axis-corrected and non-axis-corrected volume estimation was also high and might still be sufficient, depending on the degree of the required measurement accuracy.
- 28 June 2022
- Published by MDPI AG in Diagnostics
- Vol. 12 (7) , 1570
Open AccessDorsal Root Ganglia Volume—Normative Values, Correlation with Demographic Determinants and Reliability of Three Different Methods of Volumetry
Background: Dorsal root ganglia (DRG) volume assessment by MR-Neurography (MRN) has evolved to an important imaging marker in the diagnostic workup of various peripheral neuropathies and pain syndromes. The aim of this study was (1) to assess normal values of DRG volume and correlations with demographic determinants and (2) to quantify the inter-reader and inter-method reliability of three different methods of DRG volumetry. Methods: Sixty healthy subjects (mean age: 59.1, range 23–79) were examined using a 3D T2-weighted MRN of the lumbosacral plexus at 3 Tesla. Normal values of DRG L3 to S2 were obtained after exact volumetry based on manual 3D segmentation and correlations with demographic variables were assessed. For the assessment of inter-reader and inter-method reliability, DRG volumes in a subset of 25 participants were measured by two independent readers, each applying (1) exact volumetry based on 3D segmentation, (2) axis-corrected, and (3) non-axis-corrected volume estimation. Intraclass correlation coefficients were reported and the Bland–Altman analysis was conducted. Results: Mean DRG volumes ranged from 124.8 mm3 for L3 to 323.3 mm3 for S1 and did not differ between right and left DRG. DRG volume (mean of L3 to S1) correlated with body height (r = 0.42; p = 0.0008) and weight (r = 0.34; p = 0.0087). DRG of men were larger than of women (p = 0.0002); however, no difference remained after correction for body height. Inter-reader reliability was high for all three methods but best for exact volumetry (ICC = 0.99). While axis-corrected estimation was not associated with a relevant bias, non-axis-corrected estimation systematically overestimated DRG volume by on average of 15.55 mm3 (reader 1) or 18.00 mm3 (reader 2) when compared with exact volumetry. Conclusion: The here presented normal values of lumbosacral DRG volume and the correlations with height and weight may be considered in future disease specific studies and possible clinical applications. Exact volumetry was most reliable and should be considered the gold standard. However, the reliability of axis-corrected and non-axis-corrected volume estimation was also high and might still be sufficient, depending on the degree of the required measurement accuracy.