Dataset Information

Automated brain tumor identification using magnetic resonance imaging: A systematic review and meta-analysis.

ABSTRACT:

Background

Automated brain tumor identification facilitates diagnosis and treatment planning. We evaluate the performance of traditional machine learning (TML) and deep learning (DL) in brain tumor detection and segmentation, using MRI.

Methods

A systematic literature search from January 2000 to May 8, 2021 was conducted. Study quality was assessed using the Checklist for Artificial Intelligence in Medical Imaging (CLAIM). Detection meta-analysis was performed using a unified hierarchical model. Segmentation studies were evaluated using a random effects model. Sensitivity analysis was performed for externally validated studies.

Results

Of 224 studies included in the systematic review, 46 segmentation and 38 detection studies were eligible for meta-analysis. In detection, DL achieved a lower false positive rate compared to TML; 0.018 (95% CI, 0.011 to 0.028) and 0.048 (0.032 to 0.072) (P < .001), respectively. In segmentation, DL had a higher dice similarity coefficient (DSC), particularly for tumor core (TC); 0.80 (0.77 to 0.83) and 0.63 (0.56 to 0.71) (P < .001), persisting on sensitivity analysis. Both manual and automated whole tumor (WT) segmentation had "good" (DSC ≥ 0.70) performance. Manual TC segmentation was superior to automated; 0.78 (0.69 to 0.86) and 0.64 (0.53 to 0.74) (P = .014), respectively. Only 30% of studies reported external validation.

Conclusions

The comparable performance of automated to manual WT segmentation supports its integration into clinical practice. However, manual outperformance for sub-compartmental segmentation highlights the need for further development of automated methods in this area. Compared to TML, DL provided superior performance for detection and sub-compartmental segmentation. Improvements in the quality and design of studies, including external validation, are required for the interpretability and generalizability of automated models.

SUBMITTER: Kouli O

PROVIDER: S-EPMC9234754 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:(1) Background: Transfer learning refers to machine learning techniques that focus on acquiring knowledge from related tasks to improve generalization in the tasks of interest. In magnetic resonance imaging (MRI), transfer learning is important for developing strategies that address the variation in MR images from different imaging protocols or scanners. Additionally, transfer learning is beneficial for reutilizing machine learning models that were trained to solve different (but related) tasks to the task of interest. The aim of this review is to identify research directions, gaps in knowledge, applications, and widely used strategies among the transfer learning approaches applied in MR brain imaging; (2) Methods: We performed a systematic literature search for articles that applied transfer learning to MR brain imaging tasks. We screened 433 studies for their relevance, and we categorized and extracted relevant information, including task type, application, availability of labels, and machine learning methods. Furthermore, we closely examined brain MRI-specific transfer learning approaches and other methods that tackled issues relevant to medical imaging, including privacy, unseen target domains, and unlabeled data; (3) Results: We found 129 articles that applied transfer learning to MR brain imaging tasks. The most frequent applications were dementia-related classification tasks and brain tumor segmentation. The majority of articles utilized transfer learning techniques based on convolutional neural networks (CNNs). Only a few approaches utilized clearly brain MRI-specific methodology, and considered privacy issues, unseen target domains, or unlabeled data. We proposed a new categorization to group specific, widely-used approaches such as pretraining and fine-tuning CNNs; (4) Discussion: There is increasing interest in transfer learning for brain MRI. Well-known public datasets have clearly contributed to the popularity of Alzheimer's diagnostics/prognostics and tumor segmentation as applications. Likewise, the availability of pretrained CNNs has promoted their utilization. Finally, the majority of the surveyed studies did not examine in detail the interpretation of their strategies after applying transfer learning, and did not compare their approach with other transfer learning approaches.

Project description:ImportanceCovert vascular brain injury (VBI) is highly prevalent in community-dwelling older persons, but its clinical and therapeutic implications are debated.ObjectiveTo better understand the clinical significance of VBI to optimize prevention strategies for the most common age-related neurological diseases, stroke and dementia.Data sourceWe searched for articles in PubMed between 1966 and December 22, 2017, studying the association of 4 magnetic resonance imaging (MRI) markers of covert VBI (white matter hyperintensities [WMHs] of presumed vascular origin, MRI-defined covert brain infarcts [BIs], cerebral microbleeds [CMBs], and perivascular spaces [PVSs]) with incident stroke, dementia, or death.Study selectionData were taken from prospective, longitudinal cohort studies including 50 or more adults.Data extraction and synthesisWe performed inverse variance-weighted meta-analyses with random effects and z score-based meta-analyses for WMH burden. The significance threshold was P < .003 (17 independent tests). We complied with the Meta-analyses of Observational Studies in Epidemiology guidelines.Main outcomes and measuresStroke (hemorrhagic and ischemic), dementia (all and Alzheimer disease), and death.ResultsOf 2846 articles identified, 94 studies were eligible, with up to 14 529 participants for WMH, 16 012 participants for BI, 15 693 participants for CMB, and 4587 participants for PVS. Extensive WMH burden was associated with higher risk of incident stroke (hazard ratio [HR], 2.45; 95% CI, 1.93-3.12; P < .001), ischemic stroke (HR, 2.39; 95% CI, 1.65-3.47; P < .001), intracerebral hemorrhage (HR, 3.17; 95% CI, 1.54-6.52; P = .002), dementia (HR, 1.84; 95% CI, 1.40-2.43; P < .001), Alzheimer disease (HR, 1.50; 95% CI, 1.22-1.84; P < .001), and death (HR, 2.00; 95% CI, 1.69-2.36; P < .001). Presence of MRI-defined BIs was associated with higher risk of incident stroke (HR, 2.38; 95% CI, 1.87-3.04; P < .001), ischemic stroke (HR, 2.18; 95% CI, 1.67-2.85; P < .001), intracerebral hemorrhage (HR, 3.81; 95% CI, 1.75-8.27; P < .001), and death (HR, 1.64; 95% CI, 1.40-1.91; P < .001). Presence of CMBs was associated with increased risk of stroke (HR, 1.98; 95% CI, 1.55-2.53; P < .001), ischemic stroke (HR, 1.92; 95% CI, 1.40-2.63; P < .001), intracerebral hemorrhage (HR, 3.82; 95% CI, 2.15-6.80; P < .001), and death (HR, 1.53; 95% CI, 1.31-1.80; P < .001). Data on PVS were limited and insufficient to conduct meta-analyses but suggested an association of high PVS burden with increased risk of stroke, dementia, and death; this requires confirmation.Conclusions and relevanceWe report evidence that MRI markers of VBI have major clinical significance. This research prompts careful evaluation of the benefit-risk ratio for available prevention strategies in individuals with covert VBI.

Project description:OBJECTIVES:To determine prevalence and types of potentially serious incidental findings on magnetic resonance imaging (MRI) in apparently asymptomatic adults, describe factors associated with potentially serious incidental findings, and summarise information on follow-up and final diagnoses. DESIGN:Systematic review and meta-analyses. DATA SOURCES:Citation searches of relevant articles and authors' files in Medline and Embase (from inception to 25 April 2017). REVIEW METHODS:Eligible studies included prevalence and types of incidental findings detected among apparently asymptomatic adults undergoing MRI of the brain, thorax, abdomen, or brain and body. Data on study population and methods, prevalence and types of incidental findings, and final diagnoses were extracted. Pooled prevalence was estimated by random effects meta-analysis, and heterogeneity by ?2 statistics. MAIN OUTCOME MEASURES:Prevalence of potentially serious incidental findings on MRI of the brain, thorax, abdomen, and brain and body. RESULTS:Of 5905 retrieved studies, 32 (0.5%) met the inclusion criteria (n=27?643 participants). Pooled prevalence of potentially serious incidental findings was 3.9% (95% confidence interval 0.4% to 27.1%) on brain and body MRI, 1.4% (1.0% to 2.1%) on brain MRI, 1.3% (0.2% to 8.1%) on thoracic MRI, and 1.9% (0.3% to 12.0%) on abdominal MRI. Pooled prevalence rose after including incidental findings of uncertain potential seriousness (12.8% (3.9% to 34.3%), 1.7% (1.1% to 2.6%), 3.0% (0.8% to 11.3%), and 4.5% (1.5% to 12.9%), respectively). There was generally substantial heterogeneity among included studies. About half the potentially serious incidental findings were suspected malignancies (brain, 0.6% (95% confidence interval 0.4% to 0.9%); thorax, 0.6% (0.1% to 3.1%); abdomen, 1.3% (0.2% to 9.3%); brain and body, 2.3% (0.3% to 15.4%)). There were few informative data on potential sources of between-study variation or factors associated with potentially serious incidental findings. Limited data suggested that relatively few potentially serious incidental findings had serious final diagnoses (48/234, 20.5%). CONCLUSIONS:A substantial proportion of apparently asymptomatic adults will have potentially serious incidental findings on MRI, but little is known of their health consequences. Systematic, long term follow-up studies are needed to better inform on these consequences and the implications for policies on feedback of potentially serious incidental findings. SYSTEMATIC REVIEW REGISTRATION:Prospero CRD42016029472.

Project description:BackgroundThis study aimed to systematically evaluate the value of magnetic resonance imaging (MRI) and magnetic resonance arthrography (MRA) in the diagnosis of acetabular labral tears.MethodsDatabases including PubMed, Embase, Cochrane Library, Web of Science, CBM, CNKI, WanFang Data, and VIP were electronically searched to collect relevant studies on magnetic resonance in the diagnosis of acetabular labral tears from inception to September 1, 2021. Two reviewers independently screened the literature, extracted data, and assessed the risk of bias in the included studies by using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. RevMan 5.3, Meta Disc 1.4, and Stata SE 15.0 were used to investigate the diagnostic value of magnetic resonance in patients with acetabular labral tears.ResultsA total of 29 articles were included, involving 1385 participants and 1367 hips. The results of the meta-analysis showed that the pooled sensitivity, pooled specificity, pooled positive likelihood ratio, pooled negative likelihood ratio, pooled diagnostic odds ratio, area under the curve of the summary receiver operating characteristic, and Q* of MRI for diagnosing acetabular labral tears were 0.77 (95% confidence interval [CI], 0.75-0.80), 0.74 (95% CI, 0.68-0.80), 2.19 (95% CI, 1.76-2.73), 0.48 (95% CI, 0.36-0.65), 4.86 (95% CI, 3.44-6.86), 0.75, and 0.69, respectively. The pooled sensitivity, pooled specificity, pooled positive likelihood ratio, pooled negative likelihood ratio, pooled diagnostic odds ratio, area under the curve of the summary receiver operating characteristic, and Q* of MRA for diagnosing acetabular labral tears were 0.87 (95% CI, 0.84-0.89), 0.64 (95% CI, 0.57-0.71), 2.23 (95% CI, 1.57-3.16), 0.21 (95% CI, 0.16-0.27), 10.47 (95% CI, 7.09-15.48), 0.89, and 0.82, respectively.ConclusionMRI has high diagnostic efficacy for acetabular labral tears, and MRA has even higher diagnostic efficacy. Due to the limited quality and quantity of the included studies, the above results should be further validated.