Download This Paper
Preprints with The Lancet is a collaboration between The Lancet Group of journals and SSRN to facilitate the open sharing of preprints for early engagement, community comment, and collaboration. Preprints available here are not Lancet publications or necessarily under review with a Lancet journal. These preprints are early-stage research papers that have not been peer-reviewed. The usual SSRN checks and a Lancet-specific check for appropriateness and transparency have been applied. The findings should not be used for clinical or public health decision-making or presented without highlighting these facts. For more information, please see the FAQs.
Deep Learning Assisted Segmentation and Detection for Intracranial Aneurysms Magnetic Resonance T1 Imaging: Development and Validation
24 Pages Posted: 27 Jul 2022
More...Abstract
Background: There are almost no methods for IAs segmentation based on T1 images, which are the most commonly used in clinical. In this study we developed a deep learning framework for IAs segmentation and detection based on T1 images.
Methods: A retrospective, diagnostic study was carried out based on 194 IAs from the 136 patients who underwent the T1 images in Beijing Tiantan Hospital. 162 were randomly used for training and validation, and 32 were used to test the accuracy and concordance of our algorithm. We designed and assembled three convolutional neural networks to segment and detect the IAs.
Findings: Our assembled model achieved overall Dice, voxel-level sensitivity, specificity, balanced accuracy and F1 score of 0.77, 0.88, 0.999, 0.942 and 0.77, respectively. When the coincidence of the aneurysms predicted by the model with the ground truth is greater than 0.7, we consider it to be a true positive. For detection of IAs, the sensitivity reached 93.75% with 0.77 false positives per scans. At the same time, the volume segmented by our model has a high agreement and consistency with the volume labeled by experts.
Interpretation: The deep learning framework is feasible and robust for segmentation and detection of IAs. It is more in line with the actual clinical needs, and has more potential for clinical applications than DSA, CTA and MRA.
Funding Information: This work was supported by the National Natural Science Foundation of China (No. 62171300, 82171290), Natural Science Foundation of Beijing Municipality (No. 7222050, L192013), Beijing Municipal Administration of Hospital’s Ascent Plan (DFL20190501).
Declaration of Interests: All authors have no conflicts of interest to declare.
Ethics Approval Statement: This study was approved by the Institutional Ethical Committee of Beijing Tiantan Hospital Patients diagnosed with IAs between May 2015 to December 2021 were recruited.
Keywords: Intracranial Aneurysms, Detection and Segmentation, Magnetic Resonance Imaging, T1, Deep Learning
Suggested Citation: Suggested Citation