Speeding Up and Improving Image Quality in Glioblastoma MRI Protocol by Deep Learning Image Reconstruction

A fully diagnostic MRI glioma protocol is key to monitoring therapy assessment but is time-consuming and especially challenging in critically ill and uncooperative patients. Artificial intelligence demonstrated promise in reducing scan time and improving image quality simultaneously. The purpose of this study was to investigate the diagnostic performance, the impact on acquisition acceleration, and the image quality of a deep learning optimized glioma protocol of the brain. Thirty-three patients with histologically confirmed glioblastoma underwent standardized brain tumor imaging according to the glioma consensus recommendations on a 3-Tesla MRI scanner. Conventional and deep learning-reconstructed (DLR) fluid-attenuated inversion recovery, and T2- and T1-weighted contrast-enhanced Turbo spin echo images with an improved in-plane resolution, i.e., super-resolution, were acquired. Two experienced neuroradiologists independently evaluated the image datasets for subjective image quality, diagnostic confidence, tumor conspicuity, noise levels, artifacts, and sharpness. In addition, the tumor volume was measured in the image datasets according to Response Assessment in Neuro-Oncology (RANO) 2.0, as well as compared between both imaging techniques, and various clinical-pathological parameters were determined. The average time saving of DLR sequences was 30% per MRI sequence. Simultaneously, DLR sequences showed superior overall image quality (all p < 0.001), improved tumor conspicuity and image sharpness (all p < 0.001, respectively), and less image noise (all p < 0.001), while maintaining diagnostic confidence (all p > 0.05), compared to conventional images. Regarding RANO 2.0, the volume of non-enhancing non-target lesions (p = 0.963), enhancing target lesions (p = 0.993), and enhancing non-target lesions (p = 0.951) did not differ between reconstruction types. The feasibility of the deep learning-optimized glioma protocol was demonstrated with a 30% reduction in acquisition time on average and an increased in-plane resolution. The evaluated DLR sequences improved subjective image quality and maintained diagnostic accuracy in tumor detection and tumor classification according to RANO 2.0.

Evaluation of the contribution of individual arteries to the cerebral blood supply in patients with Moyamoya angiopathy: comparison of vessel-encoded arterial spin labeling and digital subtraction angiography

PURPOSE

Vessel-encoded arterial spin labeling (VE-ASL) is able to provide noninvasive information about the contribution of individual arteries to the cerebral perfusion. The aim of this study was to compare VE-ASL to the diagnostic standard digital subtraction angiography (DSA) with respect to its ability to visualize vascular territories.

METHODS

In total, 20 VE-ASL and DSA data sets of 17 patients with Moyamoya angiopathy with and without revascularization surgery were retrospectively analyzed. Two neuroradiologists independently assessed the agreement between VE-ASL and DSA using a 4-point Likert scale (no- very high agreement). Additionally, grading of the vascular supply of subterritories (A1-A2, M1-M6) on the VE-ASL images and angiograms was performed. The intermodal agreement was calculated for all subterritories in total and for the subdivision into without and after revascularization (direct or indirect bypass).

RESULTS

There was a very high agreement between the VE-ASL and the DSA data sets (median = 1, modus = 1) with a substantial inter-rater agreement (kw = 0.762 (95% CI 0.561-0.963)). The inter-modality agreement between VE-ASL and DSA in vascular subterritories was almost perfect for all subterritories (k = 0.899 (0.865-0.945)), in the subgroup of direct revascularized subterritories (k = 0.827 (0.738-0.915)), in the subgroup of indirect revascularized subterritories (k = 0.843 (0.683-1.003)), and in the subgroup of never revascularized subterritories (k = 0.958 (0.899-1.017)).

CONCLUSION

Vessel-encoded ASL seems to be a promising non-invasive method to depict the contributions of individual arteries to the cerebral perfusion before and after revascularization surgery.

Deep learning-accelerated image reconstruction in back pain-MRI imaging: reduction of acquisition time and improvement of image quality

INTRODUCTION

Low back pain is a global health issue causing disability and missed work days. Commonly used MRI scans including T1-weighted and T2-weighted images provide detailed information of the spine and surrounding tissues. Artificial intelligence showed promise in improving image quality and simultaneously reducing scan time. This study evaluates the performance of deep learning (DL)-based T2 turbo spin-echo (TSE, T2DLR) and T1 TSE (T1DLR) in lumbar spine imaging regarding acquisition time, image quality, artifact resistance, and diagnostic confidence.

MATERIAL AND METHODS

This retrospective monocentric study included 60 patients with lower back pain who underwent lumbar spinal MRI between February and April 2023. MRI parameters and DL reconstruction (DLR) techniques were utilized to acquire images. Two neuroradiologists independently evaluated image datasets based on various parameters using a 4-point Likert scale.

RESULTS

Accelerated imaging showed significantly less image noise and artifacts, as well as better image sharpness, compared to standard imaging. Overall image quality and diagnostic confidence were higher in accelerated imaging. Relevant disk herniations and spinal fractures were detected in both DLR and conventional images. Both readers favored accelerated imaging in the majority of examinations. The lumbar spine examination time was cut by 61% in accelerated imaging compared to standard imaging.

CONCLUSION

In conclusion, the utilization of deep learning-based image reconstruction techniques in lumbar spinal imaging resulted in significant time savings of up to 61% compared to standard imaging, while also improving image quality and diagnostic confidence. These findings highlight the potential of these techniques to enhance efficiency and accuracy in clinical practice for patients with lower back pain.

Deep learning-accelerated image reconstruction in MRI of the orbit to shorten acquisition time and enhance image quality

BACKGROUND AND PURPOSE

This study explores the use of deep learning (DL) techniques in MRI of the orbit to enhance imaging. Standard protocols, although detailed, have lengthy acquisition times. We investigate DL-based methods for T2-weighted and T1-weighted, fat-saturated, contrast-enhanced turbo spin echo (TSE) sequences, aiming to improve image quality, reduce acquisition time, minimize artifacts, and enhance diagnostic confidence in orbital imaging.

METHODS

In a 3-Tesla MRI study of 50 patients evaluating orbital diseases from March to July 2023, conventional (TSES ) and DL TSE sequences (TSEDL ) were used. Two neuroradiologists independently assessed the image datasets for image quality, diagnostic confidence, noise levels, artifacts, and image sharpness using a randomized and blinded 4-point Likert scale.

RESULTS

TSEDL significantly reduced image noise and artifacts, enhanced image sharpness, and decreased scan time, outperforming TSES (p < .05). TSEDL showed superior overall image quality and diagnostic confidence, with relevant findings effectively detected in both DL-based and conventional images. In 94% of cases, readers preferred accelerated imaging.

CONCLUSION

The study proved that using DL for MRI image reconstruction in orbital scans significantly cut acquisition time by 69%. This approach also enhanced image quality, reduced image noise, sharpened images, and boosted diagnostic confidence.

Targeted therapies in retinoblastoma: GD2-directed immunotherapy following autologous stem cell transplantation and evaluation of alternative target B7-H3

BACKGROUND

GD2-directed immunotherapy is highly effective in the treatment of high-risk neuroblastoma (NB), and might be an interesting target also in other high-risk tumors.

METHODS

The German-Austrian Retinoblastoma Registry, Essen, was searched for patients, who were treated with anti-GD2 monoclonal antibody (mAb) dinutuximab beta (Db) in order to evaluate toxicity, response and outcome in these patients. Additionally, we evaluated anti-GD2 antibody-dependent cell-mediated cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) in retinoblastoma cell lines in vitro. Furthermore, in vitro cytotoxicity assays directed against B7-H3 (CD276), a new identified potential target in RB, were performed.

RESULTS

We identified four patients with relapsed stage IV retinoblastoma, who were treated with Db following autologous stem cell transplantation (ASCT). Two out of two evaluable patients with detectable tumors responded to immunotherapy. One of these and another patient who received immunotherapy without residual disease relapsed 10 and 12 months after start of Db. The other patients remained in remission until last follow-up 26 and 45 months, respectively. In vitro, significant lysis of RB cell lines by ADCC and CDC with samples from patients and healthy donors and anti-GD2 and anti-CD276-mAbs were demonstrated.

CONCLUSION

Anti-GD2-directed immunotherapy represents an additional therapeutic option in high-risk metastasized RB. Moreover, CD276 is another target of interest.

Deep Learning Accelerated Image Reconstruction of Fluid-Attenuated Inversion Recovery Sequence in Brain Imaging: Reduction of Acquisition Time and Improvement of Image Quality

RATIONALE AND OBJECTIVES

Fluid-attenuated inversion recovery (FLAIR) imaging is playing an increasingly significant role in the detection of brain metastases with a concomitant increase in the number of magnetic resonance imaging (MRI) examinations. Therefore, the purpose of this study was to investigate the impact on image quality and diagnostic confidence of an innovative deep learning-based accelerated FLAIR (FLAIRDLR) sequence of the brain compared to conventional (standard) FLAIR (FLAIRS) imaging.

MATERIALS AND METHODS

Seventy consecutive patients with staging cerebral MRIs were retrospectively enrolled in this single-center study. The FLAIRDLR was conducted using the same MRI acquisition parameters as the FLAIRS sequence, except for a higher acceleration factor for parallel imaging (from 2 to 4), which resulted in a shorter acquisition time of 1:39 minute instead of 2:40 minutes (-38%). Two specialized neuroradiologists evaluated the imaging datasets using a Likert scale that ranged from 1 to 4, with 4 indicating the best score for the following parameters: sharpness, lesion demarcation, artifacts, overall image quality, and diagnostic confidence. Additionally, the image preference of the readers and the interreader agreement were assessed.

RESULTS

The average age of the patients was 63 ± 11years. FLAIRDLR exhibited significantly less image noise than FLAIRS, with P-values of< .001 and< .05, respectively. The sharpness of the images and the ability to detect lesions were rated higher in FLAIRDLR, with a median score of 4 compared to a median score of 3 in FLAIRS (P-values of<.001 for both readers). In terms of overall image quality, FLAIRDLR was rated superior to FLAIRS, with a median score of 4 vs 3 (P-values of<.001 for both readers). Both readers preferred FLAIRDLR in 68/70 cases.

CONCLUSION

The feasibility of deep learning FLAIR brain imaging was shown with additional 38% reduction in examination time compared to standard FLAIR imaging. Furthermore, this technique has shown improvement in image quality, noise reduction, and lesion demarcation.

Evaluation of the cerebrovascular reactivity in patients with Moyamoya Angiopathy by use of breath-hold fMRI: investigation of voxel-wise hemodynamic delay correction in comparison to [15O]water PET

PURPOSE

Patients with Moyamoya Angiopathy (MMA) require hemodynamic assessment to evaluate the risk of stroke. Hemodynamic evaluation by use of breath-hold-triggered fMRI (bh-fMRI) was proposed as a readily available alternative to the diagnostic standard [¹⁵O]water PET. Recent studies suggest voxel-wise hemodynamic delay correction in hypercapnia-triggered fMRI. The aim of this study was to evaluate the effect of delay correction of bh-fMRI in patients with MMA and to compare the results with [¹⁵O]water PET.

METHODS

bh-fMRI data sets of 22 patients with MMA were evaluated without and with voxel-wise delay correction within different shift ranges and compared to the corresponding [¹⁵O]water PET data sets. The effects were evaluated combined and in subgroups of data sets with most severely impaired CVR (apparent steal phenomenon), data sets with territorial time delay, and data sets with neither steal phenomenon nor delay between vascular territories.

RESULTS

The study revealed a high mean cross-correlation (r = 0.79, p < 0.001) between bh-fMRI and [¹⁵O]water PET. The correlation was strongly dependent on the choice of the shift range. Overall, no shift range revealed a significantly improved correlation between bh-fMRI and [¹⁵O]water PET compared to the correlation without delay correction. Delay correction within shift ranges with positive high high cutoff revealed a lower agreement between bh-fMRI and PET overall and in all subgroups.

CONCLUSION

Voxel-wise delay correction, in particular with shift ranges with high cutoff, should be used critically as it can lead to false-negative results in regions with impaired CVR and a lower correlation to the diagnostic standard [¹⁵O]water PET.

AB1138 ASSESSMENT OF FLUORESCENCE-OPTICAL IMAGING TECHNIQUE OF THE HANDS IN PSORIASIS AND PSORIATIC PATIENTS USING AN INNOVATIVE OBJECTIVE METHOD

Background:

Psoriasis (Pso) is one of the most common chronic inflammatory skin diseases in Europe. Psoriatic arthritis (PsA) is closely associated to Pso whereas the skin manifestation appears usually years before PsA-related symptoms emerge. Up to 30% of Pso patients develop PsA, biomarkers for its early detection are of major importance. In early PsA, changes in synovial vascularisation appear first. Imaging biomarkers for detection of changes in vascularisation might be useful for early detection of musculoskeletal disease. Fluorescence-optical imaging (FOI) is a new method to detect changes in microvascularisation of the hands. Each collected data set of the FOI system contains 360 images representing a time progression of the indocyanine green (ICG) distribution.

Objectives:

To evaluate a reader-independent assessment method for evaluation of FOI in patients with PsO and PsA.

Methods:

A prospective study including patients with dermatological confirmed skin PsO was performed. 411 patients were included from German dermatology units without PsA diagnosis but potential risk for its development. Clinical examination (CE) was performed by a qualified rheumatologist. For a reader independent evaluation of the FOI images an objective joint-based scoring method was developed. For this method, the joint areas are defined by image segmentation and scored based on generated heatmaps. To calculate a heatmap indicating conspicuous joints from a data set containing 360 images, each pixel is converted to a time series containing 360 values. From this time series, three independent values (features) are extracted: amplitude, average value and maximal slope. Thus, each pixel is reduced to three different feature values. After the three features are determined for each pixel, k-means clustering is performed on each feature. The numbers of centroids (k) are set to 3, 5, 7 and 9. 12 heatmaps (3 features à 4 ks) are calculated, which results in 12 scores for each joint as well. The clusters are then sorted dependent on their centroid value and coloured accordingly to a predefined heatmap colour palette. To finally score each joint, the pixels in the segmented joint area and their assigned cluster are summed and normalized by the area’s amount of pixels and k.

Results:

271 of the patients were investigated by the newly developed method and compared with the CE scoring. 6426 joints were labeled as healthy whereas 1162 joints were either labeled as swollen, tender or both. The result over all investigated patients for k = 9 is summed in table 1. It is observable that every average and median healthy value is lower than the corresponding affected value.

Table 1.

Resulting scores for k = 9 for all 271 patients.

Feature Statistic valueAmplitudeMeanSlopeHealthyAffectedHealthyAffectedHealthyAffectedAverage0.5030.5280.4860.5090.3950.414Median0.4960.5320.4820.5050.3890.415

Conclusion:

FOI is an innovative method that detects early changes in vascularization of the hands. So, this method can be useful in early detection of arthritis especially in risk populations such as PsO patients. The results of the objective scoring method show that a clear distinction between healthy and affected joints is possible with the average scores as well as the median values. However, if the range of the scores is considered, the overlap between healthy and affected is not neglectable. Thus, the current scoring system can be used as an indicator but not as a single classification marker. Nevertheless, the research at hand has shown the expected outcome and motivates further development on the heatmap approach.

Disclosure of Interests:

Lukas Zerweck: None declared, Ulf Henkemeier: None declared, Phuong-Ha Nguyen: None declared, Tanja Rossmanith Grant/research support from: Janssen, BMS, LEO, Pfizer, Andreas Pippow: None declared, Harald Burkhardt Grant/research support from: Pfizer, Roche, Abbvie, Consultant of: Sanofi, Pfizer, Roche, Abbvie, Boehringer Ingelheim, UCB, Eli Lilly, Chugai, Bristol Myer Scripps, Janssen, and Novartis, Speakers bureau: Sanofi, Pfizer, Roche, Abbvie, Boehringer Ingelheim, UCB, Eli Lilly, Chugai, Bristol Myer Scripps, Janssen, and Novartis, Frank Behrens Grant/research support from: Pfizer, Janssen, Chugai, Celgene, Lilly and Roche, Consultant of: Pfizer, AbbVie, Sanofi, Lilly, Novartis, Genzyme, Boehringer, Janssen, MSD, Celgene, Roche and Chugai, Michaela Köhm Grant/research support from: Pfizer, Janssen, BMS, LEO, Consultant of: BMS, Pfizer, Speakers bureau: Pfizer, BMS, Janssen, Novartis