-
A Robust Ensemble Algorithm for Ischemic Stroke Lesion Segmentation: Generalizability and Clinical Utility Beyond the ISLES Challenge
Authors:
Ezequiel de la Rosa,
Mauricio Reyes,
Sook-Lei Liew,
Alexandre Hutton,
Roland Wiest,
Johannes Kaesmacher,
Uta Hanning,
Arsany Hakim,
Richard Zubal,
Waldo Valenzuela,
David Robben,
Diana M. Sima,
Vincenzo Anania,
Arne Brys,
James A. Meakin,
Anne Mickan,
Gabriel Broocks,
Christian Heitkamp,
Shengbo Gao,
Kongming Liang,
Ziji Zhang,
Md Mahfuzur Rahman Siddiquee,
Andriy Myronenko,
Pooya Ashtari,
Sabine Van Huffel
, et al. (33 additional authors not shown)
Abstract:
Diffusion-weighted MRI (DWI) is essential for stroke diagnosis, treatment decisions, and prognosis. However, image and disease variability hinder the development of generalizable AI algorithms with clinical value. We address this gap by presenting a novel ensemble algorithm derived from the 2022 Ischemic Stroke Lesion Segmentation (ISLES) challenge. ISLES'22 provided 400 patient scans with ischemi…
▽ More
Diffusion-weighted MRI (DWI) is essential for stroke diagnosis, treatment decisions, and prognosis. However, image and disease variability hinder the development of generalizable AI algorithms with clinical value. We address this gap by presenting a novel ensemble algorithm derived from the 2022 Ischemic Stroke Lesion Segmentation (ISLES) challenge. ISLES'22 provided 400 patient scans with ischemic stroke from various medical centers, facilitating the development of a wide range of cutting-edge segmentation algorithms by the research community. Through collaboration with leading teams, we combined top-performing algorithms into an ensemble model that overcomes the limitations of individual solutions. Our ensemble model achieved superior ischemic lesion detection and segmentation accuracy on our internal test set compared to individual algorithms. This accuracy generalized well across diverse image and disease variables. Furthermore, the model excelled in extracting clinical biomarkers. Notably, in a Turing-like test, neuroradiologists consistently preferred the algorithm's segmentations over manual expert efforts, highlighting increased comprehensiveness and precision. Validation using a real-world external dataset (N=1686) confirmed the model's generalizability. The algorithm's outputs also demonstrated strong correlations with clinical scores (admission NIHSS and 90-day mRS) on par with or exceeding expert-derived results, underlining its clinical relevance. This study offers two key findings. First, we present an ensemble algorithm (https://github.com/Tabrisrei/ISLES22_Ensemble) that detects and segments ischemic stroke lesions on DWI across diverse scenarios on par with expert (neuro)radiologists. Second, we show the potential for biomedical challenge outputs to extend beyond the challenge's initial objectives, demonstrating their real-world clinical applicability.
△ Less
Submitted 3 April, 2024; v1 submitted 28 March, 2024;
originally announced March 2024.
-
Association of bone mineral density with reoperation rate following instrumented lumbar spinal fusion
Authors:
Maximilian T. Löffler,
Niklas Loreck,
Johannes Kaesmacher,
Felix Zibold,
Ehab Shiban,
Anna Rienmüller,
Martin Vazan,
Bernhard Meyer,
Yu-Mi Ryang,
Jan S. Kirschke
Abstract:
Low bone mineral density (BMD) is believed to influence the outcome of instrumented spinal surgery and can lead to reoperation. Purpose of this observational and case-control study was to investigate the association of BMD with the risk of reoperation following instrumented lumbar spinal fusion (LSF). For the observational study, 81 patients were included who received LSF with and without augmenta…
▽ More
Low bone mineral density (BMD) is believed to influence the outcome of instrumented spinal surgery and can lead to reoperation. Purpose of this observational and case-control study was to investigate the association of BMD with the risk of reoperation following instrumented lumbar spinal fusion (LSF). For the observational study, 81 patients were included who received LSF with and without augmentation. For the case-control study, 18 patients who had reoperation following LSF were matched to 26 patients who did not have reoperation (matched by sex, age +/- 5 years, fused levels and PMMA-augmentation). Opportunistic BMD screening was performed in perioperative CT scans using asynchronous calibration. Mean BMD was compared between patients with and without reoperation in augmented and non-augmented surgeries. In the observational study, prevalence of osteoporosis (BMD < 80 mg/cc) was 29% in non-augmented and 85% in augmented LSF. Seven of 48 patients with non-augmented (15%) and 4 of 33 patients with augmented LSF (12%) had reoperation. In non-augmented LSF, patients with reoperation had significantly lower BMD than patients without reoperation (P = 0.005). The best cut-off to predict reoperation after non-augmented LSF was BMD < 83.7 mg/cc. In the case-control study, patients with reoperation presented numerically lower BMD of 78.8 +/- 33.1 mg/cc than patients without reoperation with BMD of 89.4 +/- 39.7 mg/cc (P = 0.357). Despite much lower BMD surgeries with PMMA-augmentation showed no higher reoperation rate compared to non-augmented surgeries. Patients with reoperation following LSF showed slightly lower BMD compared to matched patients without reoperation, but the difference was not statistically significant. Opportunistic BMD screening can be performed in preoperative CT, thus informing about osteoporotic bone, a potential risk factor of surgery failure.
△ Less
Submitted 7 November, 2019;
originally announced November 2019.
-
Synthetic Perfusion Maps: Imaging Perfusion Deficits in DSC-MRI with Deep Learning
Authors:
Andreas Hess,
Raphael Meier,
Johannes Kaesmacher,
Simon Jung,
Fabien Scalzo,
David Liebeskind,
Roland Wiest,
Richard McKinley
Abstract:
In this work, we present a novel convolutional neural net- work based method for perfusion map generation in dynamic suscepti- bility contrast-enhanced perfusion imaging. The proposed architecture is trained end-to-end and solely relies on raw perfusion data for inference. We used a dataset of 151 acute ischemic stroke cases for evaluation. Our method generates perfusion maps that are comparable t…
▽ More
In this work, we present a novel convolutional neural net- work based method for perfusion map generation in dynamic suscepti- bility contrast-enhanced perfusion imaging. The proposed architecture is trained end-to-end and solely relies on raw perfusion data for inference. We used a dataset of 151 acute ischemic stroke cases for evaluation. Our method generates perfusion maps that are comparable to the target maps used for clinical routine, while being model-free, fast, and less noisy.
△ Less
Submitted 11 June, 2018;
originally announced June 2018.
