A clinically applicable deep-learning model for detecting intracranial aneurysm in computed tomography angiography images

Shi, Zhengyi; Chen, Miao; Schoepf, U. Joseph; Savage, Rock H.; Dargis, Danielle M.; Pan, Chengwei; Xue, Cheng; Li, Xiu Li; Xia, Shuang; Zhang, Xin; Gu, Yushen; Zhang, Yonggang; Hu, Bin; Xu, Wenjian; Zhou, Changsheng; Luo, Song; Wang, Hao; Mao, Li; Liang, Kongming; Wen, Li-Li; Zhou, Longjiang; Yu, Yizhou; Lu, Guang Ming; Zhang, Long Jiang

doi:10.1038/s41467-020-19527-w

“…The results of this study showed that the postoperative vascular stenosis rate in the DSA group was much lower than that before the surgery, which indicates that stent placement can achieve a good effect in the treatment of vascular restenosis. Such results were in line with the research conclusion of Shi et al [ 19 ]. As the treatment of cardiovascular interventions becomes more and more complex, the success of the operation will depend more on the accuracy of the guidance system, improving the success rate of cardiovascular interventions and clear long-term curative effects.…”

Section: Discussionsupporting

confidence: 93%

Digital Subtraction Angiography Image Features under the Deep Learning Algorithm in Cardiovascular Interventional Treatment and Nursing for Vascular Restenosis

Zhao

¹

,

Zeng

²

,

Li

³

et al. 2022

Computational and Mathematical Methods in Medicine

View full text Add to dashboard Cite

The objective of this study was to explore the application value of digital subtraction angiography (DSA) images optimized by deep learning algorithms in vascular restenosis patients undergoing cardiovascular intervention and their nursing efficacy. In this study, a network model for removing artifacts was constructed based on a deep algorithm. 60 patients with coronary artery restenosis were selected as the research objects, and they were randomly divided into the CTA group guided by CT angiography (CTA) and digital subtraction angiography (DSA) group, with 30 cases in each group. The antiartifact network model constructed based on the depth algorithm was applied to the images of CTA and DSA for experiments. After cardiovascular intervention and clinical pathway nursing intervention, it was found that the diameter stenosis rate in the DSA group decreased from 65.82 ± 12.9 % to 4.7 ± 1.3 % , and the area stenosis rate decreased from 88.4 ± 14.3 % to 5.4 ± 1.7 % . During the follow-up period of 3-24 months, 3 out of 46 lesions in the DSA group showed restenosis, so the restenosis rate was 6.5%, which was significantly lower than the 18.4% in the CTA group ( P < 0.05 ). In the DSA group, there was 1 case of bleeding, 0 case of hematoma, 2 cases of urinary retention, and 0 case of hypotension, so the total incidence of adverse reactions was 10%, which was significantly lower than the 30% of the CTA group ( P < 0.05 ). The high-sensitivity C-reactive protein (hs-CRP) levels of the two groups of patients were 3.58 ± 2.02 mg / L and 4.36 ± 3.11 mg / L before surgery and 3.49 ± 2.18 mg / L and 4.57 ± 3.4 mg / L after the surgery. The postoperative hs-CRP level in the CTA group was slightly lower than that before the surgery and the postoperative hs-CRP level in the DSA group was slightly higher than that before the surgery, but they were not statistically significant ( P > 0.05 ). The hs-CRP level of the DSA group before and after the surgery was slightly higher than that of the CTA group, but there was no significant difference ( P > 0.05 ). In summary, the network model based on the deep learning algorithm can remove the artifacts in DSA images and present high-quality clear images, and convolutional neural network (CNN) algorithms had a strong ability to automatically learn features in the field of medical image processing and were worthy of being widely used and popularized. In addition, the DSA-guided intervention can reduce the rate of vascular stenosis in patients, showing low probability of postoperative restenosis and adverse reactions and a good clinical effect.

show abstract

“…Our BI analyses used site heterogeneous models and the results show their power for resolving phylogenetic relationships with Elateridae and provide for an additional example that model adequacy is critical for accurate tree reconstruction in mitochondrial phylogenomics [ 34 ]. Multiple tree searches and thorough analyses aimed to test the effect of random trapping of the analysis in a local optimum [ 72 ]. In such a way, we produced the phylogeny with a much higher statistical support.…”

Section: Discussionmentioning

confidence: 99%

Click Beetle Mitogenomics with the Definition of a New Subfamily Hapatesinae from Australasia (Coleoptera: Elateridae)

Kusy

¹

,

Motyka

²

,

Bocák

³

2020

Insects

View full text Add to dashboard Cite

Elateridae is a taxon with very unstable classification and a number of conflicting phylogenetic hypotheses have been based on morphology and molecular data. We assembled eight complete mitogenomes for seven elaterid subfamilies and merged these taxa with an additional 22 elaterids and an outgroup. The structure of the newly produced mitogenomes showed a very similar arrangement with regard to all earlier published mitogenomes for the Elateridae. The maximum likelihood and Bayesian analyses indicated that Hapatesus Candèze, 1863, is a sister of Parablacinae and Pityobiinae. Therefore, Hapatesinae, a new subfamily, is proposed for the Australian genera Hapatesus (21 spp.) and Toorongus Neboiss, 1957 (4 spp.). Parablacinae, Pityobiinae, and Hapatesinae have a putative Gondwanan origin as the constituent genera are known from the Australian region (9 genera) and Neotropical region (Tibionema Solier, 1851), and only Pityobius LeConte, 1853, occurs in the Nearctic region. Another putative Gondwanan lineage, the Afrotropical Morostomatinae, forms either a serial paraphylum with the clade of Parablacinae, Pityobiinae, and Hapatesinae or is rooted in a more terminal position, but always as an independent lineage. An Eudicronychinae lineage was either recovered as a sister to Melanotini or as a deep split inside Elaterinae and we herein transfer the group to Elaterinae as Eudicronychini, a new status. The mitochondrial genomes provide a sufficient signal for the placement of most lineages, but the deep bipartitions need to be compared with phylogenomic analyses.

show abstract

“…Previous studies have evaluated DLMs for the detection of intracranial aneurysms on CTA [18,[20][21][22] and time-of-flight (TOF)-MRA [23][24][25] and investigated whether deep learning enhancement could increase the diagnostic performance of human readers [20,25]. In the study by Park et al, artificial intelligence assistance increased the detection sensitivity of radiologists and of a neurosurgeon (2-12 years of experience) for UIAs on CTA significantly from 83% to 89% [20].…”

Section: Discussionmentioning

confidence: 99%

“…Previous studies have introduced several approaches for deep learning-based detection of aneurysms on CTA [18,[20][21][22] or magnetic resonance angiography (MRA) [23][24][25], compared the accuracy of the DLM to human readers, and investigated the deep learning-augmented diagnostic performance of physicians [20,25]. However, these studies did not include patients with aSAH and focused on unruptured intracranial aneurysms (UIAs) [20,25].…”

Section: Introductionmentioning

confidence: 99%

Deep learning assistance increases the detection sensitivity of radiologists for secondary intracranial aneurysms in subarachnoid hemorrhage

Pennig

¹

,

Hoyer

²

,

Krauskopf

³

et al. 2021

View full text Add to dashboard Cite

Purpose To evaluate whether a deep learning model (DLM) could increase the detection sensitivity of radiologists for intracranial aneurysms on CT angiography (CTA) in aneurysmal subarachnoid hemorrhage (aSAH). Methods Three different DLMs were trained on CTA datasets of 68 aSAH patients with 79 aneurysms with their outputs being combined applying ensemble learning (DLM-Ens). The DLM-Ens was evaluated on an independent test set of 104 aSAH patients with 126 aneuryms (mean volume 129.2 ± 185.4 mm3, 13.0% at the posterior circulation), which were determined by two radiologists and one neurosurgeon in consensus using CTA and digital subtraction angiography scans. CTA scans of the test set were then presented to three blinded radiologists (reader 1: 13, reader 2: 4, and reader 3: 3 years of experience in diagnostic neuroradiology), who assessed them individually for aneurysms. Detection sensitivities for aneurysms of the readers with and without the assistance of the DLM were compared. Results In the test set, the detection sensitivity of the DLM-Ens (85.7%) was comparable to the radiologists (reader 1: 91.2%, reader 2: 86.5%, and reader 3: 86.5%; Fleiss κ of 0.502). DLM-assistance significantly increased the detection sensitivity (reader 1: 97.6%, reader 2: 97.6%,and reader 3: 96.0%; overall P=.024; Fleiss κ of 0.878), especially for secondary aneurysms (88.2% of the additional aneurysms provided by the DLM). Conclusion Deep learning significantly improved the detection sensitivity of radiologists for aneurysms in aSAH, especially for secondary aneurysms. It therefore represents a valuable adjunct for physicians to establish an accurate diagnosis in order to optimize patient treatment.

show abstract

A clinically applicable deep-learning model for detecting intracranial aneurysm in computed tomography angiography images

Cited by 128 publications

References 42 publications

Digital Subtraction Angiography Image Features under the Deep Learning Algorithm in Cardiovascular Interventional Treatment and Nursing for Vascular Restenosis

Digital Subtraction Angiography Image Features under the Deep Learning Algorithm in Cardiovascular Interventional Treatment and Nursing for Vascular Restenosis

Click Beetle Mitogenomics with the Definition of a New Subfamily Hapatesinae from Australasia (Coleoptera: Elateridae)

Deep learning assistance increases the detection sensitivity of radiologists for secondary intracranial aneurysms in subarachnoid hemorrhage

Contact Info

Product

Resources

About