We conducted this Systematic Review to create an overview of the currently existing Artificial Intelligence (AI) methods for Magnetic Resonance Diffusion-Weighted Imaging (DWI)/Fluid-Attenuated Inversion Recovery (FLAIR)—mismatch assessment and to determine how well DWI/FLAIR mismatch algorithms perform compared to domain experts. We searched PubMed Medline, Ovid Embase, Scopus, Web of Science, Cochrane, and IEEE Xplore literature databases for relevant studies published between 1 January 2017 and 20 November 2022, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. We assessed the included studies using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. Five studies fit the scope of this review. The area under the curve ranged from 0.74 to 0.90. The sensitivity and specificity ranged from 0.70 to 0.85 and 0.74 to 0.84, respectively. Negative predictive value, positive predictive value, and accuracy ranged from 0.55 to 0.82, 0.74 to 0.91, and 0.73 to 0.83, respectively. In a binary classification of ±4.5 h from stroke onset, the surveyed AI methods performed equivalent to or even better than domain experts. However, using the relation between time since stroke onset (TSS) and increasing visibility of FLAIR hyperintensity lesions is not recommended for the determination of TSS within the first 4.5 h. An AI algorithm on DWI/FLAIR mismatch assessment focused on treatment eligibility, outcome prediction, and consideration of patient-specific data could potentially increase the proportion of stroke patients with unknown onset who could be treated with thrombolysis.

We conducted a systematic review of the current status of machine learning (ML) algorithms’ ability to identify multiple brain diseases, and we evaluated their applicability for improving existing scan acquisition and interpretation workflows. PubMed Medline, Ovid Embase, Scopus, Web of Science, and IEEE Xplore literature databases were searched for relevant studies published between January 2017 and February 2022. The quality of the included studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 tool. The applicability of ML algorithms for successful workflow improvement was qualitatively assessed based on the satisfaction of three clinical requirements. A total of 19 studies were included for qualitative synthesis. The included studies performed classification tasks (n = 12) and segmentation tasks (n = 7). For classification algorithms, the area under the receiver operating characteristic curve (AUC) ranged from 0.765 to 0.997, while accuracy, sensitivity, and specificity ranged from 80% to 100%, 72% to 100%, and 65% to 100%, respectively. For segmentation algorithms, the Dice coefficient ranged from 0.300 to 0.912. No studies satisfied all clinical requirements for successful workflow improvements due to key limitations pertaining to the study’s design, study data, reference standards, and performance reporting. Standardized reporting guidelines tailored for ML in radiology, prospective study designs, and multi-site testing could help alleviate this.

To evaluate the performance of an AI tool and relevant radiology professionals in detecting brain infarcts, intracranial hemorrhages, and tumors using abbreviated brain MRI scan protocols as prerequisite for an AI-driven workflow that dynamically selects additional imaging sequences based on real-time imaging findings. A retrospective, consecutively enriched cohort of routine adult brain MRI scans from four Danish hospitals was constructed. Three consultant neuroradiologists, three radiology residents, three MR technologists, and an AI tool detected brain infarcts, hemorrhages, and tumors using an abbreviated 3-sequence protocol (DWI, SWI/T2*-GRE, T2-FLAIR) or 4-sequence protocol (DWI, SWI/T2*-GRE, T2-FLAIR, T1W) in a non-overlapping three-way split cross-over design. Reference findings were established from radiological reports and independent image reviews. A total of 414 patients were included (57 years ± 19, 238 men) with 65 confirmed brain infarcts, 65 hemorrhages, and 65 tumors. The AI tool achieved detection sensitivities of 94 % (61 of 65 scans, 95 %CI:85-98 %), 82 % (53 of 65 scans, 95 %CI:70-90 %), and 74 % (48 of 65 scans, 95 %CI:61-84 %) for brain infarcts, hemorrhages, and tumors, respectively, and corresponding specificities of 86 % (300 of 349 scans, 95 %CI:82-89 %), 84 % (293 of 349 scans, 95 %CI:80-88 %), and 62 % (217 of 349 scans, 95 %CI:57-67 %). The tool achieved sensitivities comparable to those of neuroradiologists using abbreviated scan protocols, and it surpassed MR technologists in tumor detection, though its specificities were lower. The initial steps towards automated real-time scan adaptation are feasible and can potentially improve detection sensitivity, balancing reduced specificities compared to the current standard of care.

Background: Selecting optimal brain magnetic resonance imaging (MRI) protocols is a manual, error-prone process, often complicated by incomplete clinical information. Automated AI-based analysis of initial imaging sequences offers a potential strategy for dynamic protocol adaptation while the patient is still in the scanner. Purpose: To evaluate the diagnostic performance and agreement with neuroradiologists of an AI tool designed to automatically adapt brain MRI protocols based on the detection of critical findings (brain infarcts, acute hemorrhages, and mass lesions) using three initial imaging sequences. Material and Methods: We retrospectively collected consecutive cohorts of brain MRI scans from two tertiary medical centers. The cohorts were consecutively enriched with positive findings of brain infarcts, hemorrhages, and mass lesions. An AI tool and neuroradiologists independently assessed three sequences (diffusion-weighted imaging, T2-FLAIR, SWI/T2*-GRE) for critical findings and recommended protocol adaptations from seven options. Diagnostic performance was compared against reference findings based on radiological reports and de novo imaging review. Results: A total of 752 patients were included (325 men; mean age=61 years). The AI tool's pooled sensitivity for detecting infarcts, hemorrhages, and mass lesions was 92% (95% CI=86–96), 75% (95% CI=64–84), and 71% (95% CI=61–79), with pooled specificity of 93% (95% CI=90–95), 86% (95% CI=83–88), and 90% (95% CI=87–92), respectively. Agreement on protocol adaptation between the AI and neuroradiologists was moderate (κ=0.47), though concordance was high (84%–87%) for scans requiring no further adaptations. Conclusion: The AI tool demonstrated reasonable pathology detection, relevant protocol recommendations, and potential to ensure appropriate imaging protocols in high-volume, low-risk scan scenarios, but expert oversight is required.