Comprehensive Review of the Utility of Dynamic Contrast-Enhanced MRI for the Diagnosis and Treatment Assessment of Spinal Benign and Malignant Osseous Disease.

Conventional MRI is currently the preferred imaging technique for detection and evaluation of malignant spinal lesions. However, this technique is limited in its ability to assess tumor viability. Unlike conventional MRI, dynamic contrast-enhanced (DCE) MRI provides insight into the physiologic and hemodynamic characteristics of malignant spinal tumors and has been utilized in different types of spinal diseases. DCE has been shown to be especially useful in the cancer setting; specifically, DCE can discriminate between malignant and benign vertebral compression fractures as well as between atypical hemangiomas and metastases. DCE has also been shown to differentiate between different types of metastases. Furthermore, DCE can be useful in the assessment of radiation therapy for spinal metastases, including the prediction of tumor recurrence. This review considers data analysis methods utilized in prior studies of DCE-MRI data acquisition and clinical implications.

Foundational Segmentation Models and Clinical Data Mining Enable Accurate Computer Vision for Lung Cancer.

This study aims to assess the effectiveness of integrating Segment Anything Model (SAM) and its variant MedSAM into the automated mining, object detection, and segmentation (MODS) methodology for developing robust lung cancer detection and segmentation models without post hoc labeling of training images. In a retrospective analysis, 10,000 chest computed tomography scans from patients with lung cancer were mined. Line measurement annotations were converted to bounding boxes, excluding boxes < 1 cm or > 7 cm. The You Only Look Once object detection architecture was used for teacher-student learning to label unannotated lesions on the training images. Subsequently, a final tumor detection model was trained and employed with SAM and MedSAM for tumor segmentation. Model performance was assessed on a manually annotated test dataset, with additional evaluations conducted on an external lung cancer dataset before and after detection model fine-tuning. Bootstrap resampling was used to calculate 95% confidence intervals. Data mining yielded 10,789 line annotations, resulting in 5403 training boxes. The baseline detection model achieved an internal F1 score of 0.847, improving to 0.860 after self-labeling. Tumor segmentation using the final detection model attained internal Dice similarity coefficients (DSCs) of 0.842 (SAM) and 0.822 (MedSAM). After fine-tuning, external validation showed an F1 of 0.832 and DSCs of 0.802 (SAM) and 0.804 (MedSAM). Integrating foundational segmentation models into the MODS framework results in high-performing lung cancer detection and segmentation models using only mined clinical data. Both SAM and MedSAM hold promise as foundational segmentation models for radiology images.

Longitudinal Evaluation of DCE-MRI as an Early Indicator of Progression after Standard Therapy in Glioblastoma.

Background and Purpose: Distinguishing treatment-induced imaging changes from progressive disease has important implications for avoiding inappropriate discontinuation of a treatment. Our goal in this study is to evaluate the utility of dynamic contrast-enhanced (DCE) perfusion MRI as a biomarker for the early detection of progression. We hypothesize that DCE-MRI may have the potential as an early predictor for the progression of disease in GBM patients when compared to the current standard of conventional MRI. Methods: We identified 26 patients from 2011 to 2023 with newly diagnosed primary glioblastoma by histopathology and gross or subtotal resection of the tumor. Then, we classified them into two groups: patients with progression of disease (POD) confirmed by pathology or change in chemotherapy and patients with stable disease without evidence of progression or need for therapy change. Finally, at least three DCE-MRI scans were performed prior to POD for the progression cohort, and three consecutive DCE-MRI scans were performed for those with stable disease. The volume of interest (VOI) was delineated by a neuroradiologist to measure the maximum values for Ktrans and plasma volume (Vp). A Friedman test was conducted to evaluate the statistical significance of the parameter changes between scans. Results: The mean interval between subsequent scans was 57.94 days, with POD-1 representing the first scan prior to POD and POD-3 representing the third scan. The normalized maximum Vp values for POD-3, POD-2, and POD-1 are 1.40, 1.86, and 3.24, respectively (FS = 18.00, p = 0.0001). It demonstrates that Vp max values are progressively increasing in the three scans prior to POD when measured by routine MRI scans. The normalized maximum Ktrans values for POD-1, POD-2, and POD-3 are 0.51, 0.09, and 0.51, respectively (FS = 1.13, p < 0.57). Conclusions: Our analysis of the longitudinal scans leading up to POD significantly correlated with increasing plasma volume (Vp). A longitudinal study for tumor perfusion change demonstrated that DCE perfusion could be utilized as an early predictor of tumor progression.

Differentiating Low-Grade from High-Grade Intracranial Ependymomas: Comparison of Dynamic Contrast-Enhanced MRI and Diffusion-Weighted Imaging.

BACKGROUND AND PURPOSE: The aim of this study was to determine the diagnostic value of fractional plasma volume derived from dynamic contrast-enhanced perfusion MR imaging versus ADC, obtained from DWI in differentiating between grade 2 (low-grade) and grade 3 (high-grade) intracranial ependymomas. MATERIALS AND METHODS: A hospital database was created for the period from January 2013 through June 2022, including patients with histologically-proved ependymoma diagnosis with available dynamic contrast-enhanced MR imaging. Both dynamic contrast-enhanced perfusion and DWI were performed on each patient using 1.5T and 3T scanners. Fractional plasma volume maps and ADC maps were calculated. ROIs were defined by a senior neuroradiologist manually by including the enhancing tumor on every section and conforming a VOI to obtain the maximum value of fractional plasma volume (Vpmax) and the minimum value of ADC (ADCmin). A Mann-Whitney U test at a significance level of corrected P = .01 was used to evaluate the differences. Additionally, receiver operating characteristic curve analysis was applied to assess the sensitivity and specificity of Vpmax and ADCmin values. RESULTS: A total of 20 patients with ependymomas (10 grade 2 tumors and 10 grade 3 tumors) were included. Vpmax values for grade 3 ependymomas were significantly higher (P < .002) than those for grade 2. ADCmin values were overall lower in high-grade lesions. However, no statistically significant differences were found (P = .12114). CONCLUSIONS: As a dynamic contrast-enhanced perfusion MR imaging metric, fractional plasma volume can be used as an indicator to differentiate grade 2 and grade 3 ependymomas. Dynamic contrast-enhanced perfusion MR imaging plays an important role with high diagnostic value in differentiating low- and high-grade ependymoma.

Tumors Affect the Metabolic Connectivity of the Human Brain Measured by 18 F-FDG PET.

PURPOSE: 18 F-FDG PET captures the relationship between glucose metabolism and synaptic activity, allowing for modeling brain function through metabolic connectivity. We investigated tumor-induced modifications of brain metabolic connectivity. PATIENTS AND METHODS: Forty-three patients with left hemispheric tumors and 18 F-FDG PET/MRI were retrospectively recruited. We included 37 healthy controls (HCs) from the database CERMEP-IDB-MRXFDG. We analyzed the whole brain and right versus left hemispheres connectivity in patients and HC, frontal versus temporal tumors, active tumors versus radiation necrosis, and patients with high Karnofsky performance score (KPS = 100) versus low KPS (KPS < 70). Results were compared with 2-sided t test ( P < 0.05). RESULTS: Twenty high-grade glioma, 4 low-grade glioma, and 19 metastases were included. The patients' whole-brain network displayed lower connectivity metrics compared with HC ( P < 0.001), except assortativity and betweenness centrality ( P = 0.001). The patients' left hemispheres showed decreased similarity, and lower connectivity metrics compared with the right ( P < 0.01), with the exception of betweenness centrality ( P = 0.002). HC did not show significant hemispheric differences. Frontal tumors showed higher connectivity metrics ( P < 0.001) than temporal tumors, but lower betweenness centrality ( P = 4.5 -7 ). Patients with high KPS showed higher distance local efficiency ( P = 0.01), rich club coefficient ( P = 0.0048), clustering coefficient ( P = 0.00032), betweenness centrality ( P = 0.008), and similarity ( P = 0.0027) compared with low KPS. Patients with active tumor(s) (14/43) demonstrated significantly lower connectivity metrics compared with necroses. CONCLUSIONS: Tumors cause reorganization of metabolic brain networks, characterized by formation of new connections and decreased centrality. Patients with frontal tumors retained a more efficient, centralized, and segregated network than patients with temporal tumors. Stronger metabolic connectivity was associated with higher KPS.

T1-Weighted, Dynamic Contrast-Enhanced MR Perfusion Imaging Can Differentiate between Treatment Success and Failure in Spine Metastases Undergoing Radiation Therapy.

BACKGROUND AND PURPOSE: Current imaging techniques have difficulty differentiating treatment success and failure in spinal metastases undergoing radiation therapy. This study investigated the correlation between changes in dynamic contrast-enhanced MR imaging perfusion parameters and clinical outcomes following radiation therapy for spinal metastases. We hypothesized that perfusion parameters will outperform traditional size measurements in discriminating treatment success and failure. MATERIALS AND METHODS: This retrospective study included 49 patients (mean age, 63 [SD, 13] years; 29 men) with metastatic lesions treated with radiation therapy who underwent dynamic contrast-enhanced MR imaging. The median time between radiation therapy and follow-up dynamic contrast-enhanced MR imaging was 62 days. We divided patients into 2 groups: clinical success (n = 38) and failure (n = 11). Failure was defined as PET recurrence (n = 5), biopsy-proved (n = 1) recurrence, or an increase in tumor size (n = 7), while their absence defined clinical success. A Mann-Whitney U test was performed to assess differences between groups. RESULTS: The reduction in plasma volume was greater in the success group than in the failure group (-57.3% versus +88.2%, respectively; P < .001). When we assessed the success of treatment, the sensitivity of plasma volume was 91% (10 of 11; 95% CI, 82%-97%) and the specificity was 87% (33 of 38; 95% CI, 73%-94%). The sensitivity of size measurements was 82% (9 of 11; 95% CI, 67%-90%) and the specificity was 47% (18 of 38; 95% CI, 37%-67%). CONCLUSIONS: The specificity of plasma volume was higher than that of conventional size measurements, suggesting that dynamic contrast-enhanced MR imaging is a powerful tool to discriminate between treatment success and failure.

Association of Lack of Speech Arrest During Cortical Stimulation With Interhemispheric Reorganization of the Functional Language Network in Patients With Brain Tumors.

BACKGROUND. Brain tumors induce language reorganization, which may influence the extent of resection in surgical planning. Direct cortical stimulation (DCS) allows definitive language mapping during awake surgery by locating areas of speech arrest (SA) surrounding the tumor. Although functional MRI (fMRI) combined with graph theory analysis can illustrate whole-brain network reorganization, few studies have corroborated these findings with DCS intraoperative mapping and clinical language performance. OBJECTIVE. We evaluated whether patients with low-grade gliomas (LGGs) without SA during DCS show increased right-hemispheric connections and better speech performance compared with patients with SA. METHODS. We retrospectively recruited 44 consecutive patients with left perisylvian LGG, preoperative language task-based fMRI, speech performance evaluation, and awake surgery with DCS. We generated language networks from ROIs corresponding to known language areas (i.e., language core) on fMRI using optimal percolation. Language core connectivity in the left and right hemispheres was quantified as fMRI laterality index (LI) and connectivity LI on the basis of fMRI activation maps and connectivity matrices. We compared fMRI LI and connectivity LI between patients with SA and without SA and used multivariable logistic regression (p < .05) to assess associations between DCS and connectivity LI, fMRI LI, tumor location, Broca area and Wernicke area involvement, prior treatments, age, handedness, sex, tumor size, and speech deficit before surgery, within 1 week after surgery, and 3-6 months after surgery. RESULTS. Patients with SA showed left-dominant connectivity; patients without SA lateralized more to the right hemisphere (p < .001). Between patients with SA and those without, fMRI LI was not significantly different. Patients without SA showed right-greater-than-left connectivity of Broca area and premotor area compared with patients with SA. Regression analysis showed significant association between no SA and right-lateralized connectivity LI (p < .001) and fewer speech deficits before (p < .001) and 1 week after (p = .02) surgery. CONCLUSION. Patients without SA had increased right-hemispheric connections and right translocation of the language core, suggesting language reorganization. Lack of interoperative SA was associated with fewer speech deficits both before and immediately after surgery. CLINICAL IMPACT. These findings support tumor-induced language plasticity as a compensatory mechanism, which may lead to fewer postsurgical deficits and allow extended resection.

Dynamic Contrast-Enhanced MR Perfusion: Role in Diagnosis and Treatment Follow-Up in Patients with Vertebral Body Tumors.

Recent therapeutic advances have led to increased survival times for patients with metastatic disease. Key to survival is early diagnosis and subsequent treatment as well as early detection of treatment failure allowing for therapy modifications. Conventional MR imaging techniques of the spine can be at times suboptimal for identifying viable tumor, as structural changes and imaging characteristics may not differ pretreatment and posttreatment. Advanced imaging techniques such as DCE-MRI can allow earlier and more accurate noninvasive assessment of viable disease by characterizing physiologic changes and tumor microvasculature.

Magnetic resonance relaxometry in quantitative imaging of brain gliomas: A literature review.

Magnetic resonance (MR) relaxometry is a quantitative imaging method that measures tissue relaxation properties. This review discusses the state of the art of clinical proton MR relaxometry for glial brain tumors. Current MR relaxometry technology also includes MR fingerprinting and synthetic MRI, which solve the inefficiencies and challenges of earlier techniques. Despite mixed results regarding its capability for brain tumor differential diagnosis, there is growing evidence that MR relaxometry can differentiate between gliomas and metastases and between glioma grades. Studies of the peritumoral zones have demonstrated their heterogeneity and possible directions of tumor infiltration. In addition, relaxometry offers T2* mapping that can define areas of tissue hypoxia not discriminated by perfusion assessment. Studies of tumor therapy response have demonstrated an association between survival and progression terms and dynamics of native and contrast-enhanced tumor relaxometric profiles. In conclusion, MR relaxometry is a promising technique for glial tumor diagnosis, particularly in association with neuropathological studies and other imaging techniques.

Delay of Aortic Arterial Input Function Time Improves Detection of Malignant Vertebral Body Lesions on Dynamic Contrast-Enhanced MRI Perfusion.

Dynamic contrast-enhanced MRI (DCE) is an emerging modality in the study of vertebral body malignancies. DCE-MRI analysis relies on a pharmacokinetic model, which assumes that contrast uptake is simultaneous in the feeding of arteries and tissues of interest. While true in the highly vascularized brain, the perfusion of the spine is delayed. This delay of contrast reaching vertebral body lesions can affect DCE-MRI analyses, leading to misdiagnosis for the presence of active malignancy in the bone marrow. To overcome the limitation of delayed contrast arrival to vertebral body lesions, we shifted the arterial input function (AIF) curve over a series of phases and recalculated the plasma volume values (Vp) for each phase shift. We hypothesized that shifting the AIF tracer curve would better reflect actual contrast perfusion, thereby improving the accuracy of Vp maps in metastases. We evaluated 18 biopsy-proven vertebral body metastases in which standard DCE-MRI analysis failed to demonstrate the expected increase in Vp. We manually delayed the AIF curve for multiple phases, defined as the scan-specific phase temporal resolution, and analyzed DCE-MRI parameters with the new AIF curves. All patients were found to require at least one phase-shift delay in the calculated AIF to better visualize metastatic spinal lesions and improve quantitation of Vp. Average normalized Vp values were 1.78 ± 1.88 for zero phase shifts (P0), 4.72 ± 4.31 for one phase shift (P1), and 5.59 ± 4.41 for two phase shifts (P2). Mann-Whitney U tests obtained p-values = 0.003 between P0 and P1, and 0.0004 between P0 and P2. This study demonstrates that image processing analysis for DCE-MRI in patients with spinal metastases requires a careful review of signal intensity curve, as well as a possible adjustment of the phase of aortic AIF to increase the accuracy of Vp.

Fast, light, and scalable: harnessing data-mined line annotations for automated tumor segmentation on brain MRI.

OBJECTIVES: While fully supervised learning can yield high-performing segmentation models, the effort required to manually segment large training sets limits practical utility. We investigate whether data mined line annotations can facilitate brain MRI tumor segmentation model development without requiring manually segmented training data. METHODS: In this retrospective study, a tumor detection model trained using clinical line annotations mined from PACS was leveraged with unsupervised segmentation to generate pseudo-masks of enhancing tumors on T1-weighted post-contrast images (9911 image slices; 3449 adult patients). Baseline segmentation models were trained and employed within a semi-supervised learning (SSL) framework to refine the pseudo-masks. Following each self-refinement cycle, a new model was trained and tested on a held-out set of 319 manually segmented image slices (93 adult patients), with the SSL cycles continuing until Dice score coefficient (DSC) peaked. DSCs were compared using bootstrap resampling. Utilizing the best-performing models, two inference methods were compared: (1) conventional full-image segmentation, and (2) a hybrid method augmenting full-image segmentation with detection plus image patch segmentation. RESULTS: Baseline segmentation models achieved DSC of 0.768 (U-Net), 0.831 (Mask R-CNN), and 0.838 (HRNet), improving with self-refinement to 0.798, 0.871, and 0.873 (each p < 0.001), respectively. Hybrid inference outperformed full image segmentation alone: DSC 0.884 (Mask R-CNN) vs. 0.873 (HRNet), p < 0.001. CONCLUSIONS: Line annotations mined from PACS can be harnessed within an automated pipeline to produce accurate brain MRI tumor segmentation models without manually segmented training data, providing a mechanism to rapidly establish tumor segmentation capabilities across radiology modalities. KEY POINTS: • A brain MRI tumor detection model trained using clinical line measurement annotations mined from PACS was leveraged to automatically generate tumor segmentation pseudo-masks. • An iterative self-refinement process automatically improved pseudo-mask quality, with the best-performing segmentation pipeline achieving a Dice score of 0.884 on a held-out test set. • Tumor line measurement annotations generated in routine clinical radiology practice can be harnessed to develop high-performing segmentation models without manually segmented training data, providing a mechanism to rapidly establish tumor segmentation capabilities across radiology modalities.

Functional MRI during tongue strength tasks before and after partial glossectomy: Insights into the cortical activation of tongue motor function.

AIM: Because the tongue is a midline structure, studies on the neural correlates of lateralized tongue function are challenging and remain limited. Patients with tongue cancer who undergo unilateral partial glossectomy may be a unique cohort to study tongue-associated cortical activation, particularly regarding brain hemispheric lateralization. This longitudinal functional magnetic resonance imaging (fMRI) study investigated cortical activation changes for three tongue tasks before and after left-sided partial glossectomy in patients with squamous cell carcinoma of the tongue. METHODS: Seven patients with squamous cell carcinoma involving the left tongue who underwent fMRI before and 6 months after unilateral partial glossectomy were studied. Post-surgical changes in laterality index (LI) values for tongue-associated precentral and postcentral gyri fMRI activation were calculated for the dry swallow, tongue press, and saliva sucking tasks. Group analysis fMRI activation maps were generated for each of the three tasks. RESULTS: There were significant differences in changes in LI values post-surgery between the tongue press (p < 0.005; median: +0.24), saliva sucking (-0.10), and dry swallow tasks (-0.16). Decreased contralateral activation (change in LI ≥+0.20) was observed post-surgery during tongue press in six of seven patients, but only in two patients during saliva sucking and one patient during dry swallow (p < 0.05). There was also increased activation in the supplementary motor area following surgery. CONCLUSION: Post-surgical fMRI changes following left-sided partial glossectomy may suggest task-specific sensitivities to cortical activation changes following unilateral tongue deficits that may reflect the impacts of surgery and adaptive responses to tongue impairment.

Longitudinal Evaluation of Brain Plasticity in Low-Grade Gliomas: fMRI and Graph-Theory Provide Insights on Language Reorganization.

Language reorganization may represent an adaptive phenomenon to compensate tumor invasion of the dominant hemisphere. However, the functional changes over time underlying language plasticity remain unknown. We evaluated language function in patients with low-grade glioma (LGG), using task-based functional MRI (tb-fMRI), graph-theory and standardized language assessment. We hypothesized that functional networks obtained from tb-fMRI would show connectivity changes over time, with increased right-hemispheric participation. We recruited five right-handed patients (4M, mean age 47.6Y) with left-hemispheric LGG. Tb-fMRI and language assessment were conducted pre-operatively (pre-op), and post-operatively: post-op1 (4-8 months), post-op2 (10-14 months) and post-op3 (16-23 months). We computed the individual functional networks applying optimal percolation thresholding. Language dominance and hemispheric connectivity were quantified by laterality indices (LI) on fMRI maps and connectivity matrices. A fixed linear mixed model was used to assess the intra-patient correlation trend of LI values over time and their correlation with language performance. Individual networks showed increased inter-hemispheric and right-sided connectivity involving language areas homologues. Two patterns of language reorganization emerged: Three/five patients demonstrated a left-to-codominant shift from pre-op to post-op3 (type 1). Two/five patients started as atypical dominant at pre-op, and remained unchanged at post-op3 (type 2). LI obtained from tb-fMRI showed a significant left-to-right trend in all patients across timepoints. There were no significant changes in language performance over time. Type 1 language reorganization may be related to the treatment, while type 2 may be tumor-induced, since it was already present at pre-op. Increased inter-hemispheric and right-side connectivity may represent the initial step to develop functional plasticity.

Language reorganization in patients with left-hemispheric gliomas is associated with increased cortical volume in language-related areas and in the default mode network.

BACKGROUND: Language function may reorganize to overcome focal impairment; however, the relation between functional and structural changes in patients with brain tumors remains unclear. We investigated the cortical volume of atypical language dominant (AD) patients with left frontal-insular high-grade (HGG) and low-grade glioma (LGG). We hypothesized atypical language being associated with areas of increased cortical volume in the right hemisphere, including language areas homologues. METHODS: Patient were recruited following the criteria: left frontal-insular glioma; functional MRI and 3DT1-weighted images; no artifacts. We calculated an hemispheric language laterality index (LI), defined as: AD if LI < .2; left-dominant (LD) if LI ≥ .2. We measured cortical volume in three voxel-based morphometry (VBM) analyses: total AD vs. LD patients; AD vs. LD in HGG; AD vs. LD in LGG. We repeated the analysis in AD vs. LD healthy controls (HC). A minimum threshold of t > 2 and corrected p < .025 (Bonferroni) was employed. RESULTS: We recruited 119 patients (44LGG, 75HGG). Hemispheric LI demonstrated 64/119AD and 55/119LD patients. The first VBM analysis demonstrated significantly increased cortical volume in AD patients in the right inferior frontal gyrus (IFG), right superior temporal gyrus (STG), right insula, right fusiform gyrus (FG), right precentral gyrus, right temporal-parietal junction, right posterior cingulate cortex (PCC), right hippocampus, right- and left cerebellum. AD patients with HGG showed the same areas of significantly increased cortical volume. AD patients with LGG displayed significantly increased cortical volume in right IFG, right STG, right insula, right FG, right anterior cingulate cortex, right PCC, right dorsal-lateral prefrontal cortex. HC showed no significant results. CONCLUSION: Right-sided (atypical) language activations in patients with left-hemispheric gliomas are associated with areas of increased cortical volume. Additionally, default mode network nodes showed greater cortical volume in AD patients regardless of the tumor grade, supporting the idea of these cortices participating in the development of language plasticity.

Sensitivity of three-dimensional time-of-flight 3.0 ​T magnetic resonance angiography in visualizing the number and course of lenticulostriate arteries in patients with insular gliomas.

Background: Neurosurgical resection of insular gliomas is complicated by the possibility of iatrogenic injury to the lenticulostriate arteries (LSAs) and is associated with devastating neurological complications, hence the need to accurately assess the number of LSAs and their relationship to the tumor preoperatively. Methods: The study included 24 patients with insular gliomas who underwent preoperative 3D-TOF MRA to visualize LSAs. The agreement of preoperative magnetic resonance imaging with intraoperative data in terms of the number of LSAs and their invasion by the tumor was assessed using the Kendall rank correlation coefficient and Cohen's Kappa with linear weighting. Agreement between experts performing image analysis was estimated using Cohen's Kappa with linear weighting. Results: The number of LSAs arising from the M1 segment varied from 0 to 9 (mean 4.3 â€‹± â€‹0.37) as determined by 3D-TOF MRA and 2-6 (mean 4.25 â€‹± â€‹0.25) as determined intraoperatively, κ â€‹= â€‹0.51 (95% CI: 0.25-0.76) and τ â€‹= â€‹0.64 (p â€‹< â€‹0.001). LSAs were encased by the tumor in 11 patients (confirmed intraoperatively in 9 patients). LSAs were displaced medially in 8 patients (confirmed intraoperatively in 8 patients). The tumor partially involved the LSAs and displaced them in 5 patients (confirmed intraoperatively in 7 patients), κ â€‹= â€‹0.87 (95% CI: 0.70-1), τ â€‹= â€‹0.93 (p â€‹< â€‹0.001). 3D-TOF MRA demonstrated high sensitivity (100%, 95% CI: 0.63-1) and high specificity (86.67%, 95% CI: 0.58-0.98) in determining the LSA-tumor interface. Conclusions: 3D-TOF MRA at 3T demonstrated sensitivity in determining the LSA-tumor interface and the number of LSAs in patients with insular gliomas.

Brain Functional Connectivity in Low- and High-Grade Gliomas: Differences in Network Dynamics Associated with Tumor Grade and Location.

Brain tumors lead to modifications of brain networks. Graph theory plays an important role in clarifying the principles of brain connectivity. Our objective was to investigate network modifications related to tumor grade and location using resting-state functional magnetic resonance imaging (fMRI) and graph theory. We retrospectively studied 30 low-grade (LGG), 30 high-grade (HGG) left-hemispheric glioma patients and 20 healthy controls (HC) with rs-fMRI. Tumor location was labeled as: frontal, temporal, parietal, insular or occipital. We collected patients' clinical data from records. We analyzed whole-brain and hemispheric networks in all patients and HC. Subsequently, we studied lobar networks in subgroups of patients divided by tumor location. Seven graph-theoretical metrics were calculated (FDR p < 0.05). Connectograms were computed for significant nodes. The two-tailed Student t-test or Mann−Whitney U-test (p < 0.05) were used to compare graph metrics and clinical data. The hemispheric network analysis showed increased ipsilateral connectivity for LGG (global efficiency p = 0.03) and decreased contralateral connectivity for HGG (degree/cost p = 0.028). Frontal and temporal tumors showed bilateral modifications; parietal and insular tumors showed only local effects. Temporal tumors led to a bilateral decrease in all graph metrics. Tumor grade and location influence the pattern of network reorganization. LGG may show more favorable network changes than HGG, reflecting fewer clinical deficits.

Deep Learning Achieves Neuroradiologist-Level Performance in Detecting Hydrocephalus Requiring Treatment.

In large clinical centers a small subset of patients present with hydrocephalus that requires surgical treatment. We aimed to develop a screening tool to detect such cases from the head MRI with performance comparable to neuroradiologists. We leveraged 496 clinical MRI exams collected retrospectively at a single clinical site from patients referred for any reason. This diagnostic dataset was enriched to have 259 hydrocephalus cases. A 3D convolutional neural network was trained on 16 manually segmented exams (ten hydrocephalus) and subsequently used to automatically segment the remaining 480 exams and extract volumetric anatomical features. A linear classifier of these features was trained on 240 exams to detect cases of hydrocephalus that required treatment with surgical intervention. Performance was compared to four neuroradiologists on the remaining 240 exams. Performance was also evaluated on a separate screening dataset of 451 exams collected from a routine clinical population to predict the consensus reading from four neuroradiologists using images alone. The pipeline was also tested on an external dataset of 31 exams from a 2nd clinical site. The most discriminant features were the Magnetic Resonance Hydrocephalic Index (MRHI), ventricle volume, and the ratio between ventricle and brain volume. At matching sensitivity, the specificity of the machine and the neuroradiologists did not show significant differences for detection of hydrocephalus on either dataset (proportions test, p > 0.05). ROC performance compared favorably with the state-of-the-art (AUC 0.90-0.96), and replicated in the external validation. Hydrocephalus cases requiring treatment can be detected automatically from MRI in a heterogeneous patient population based on quantitative characterization of brain anatomy with performance comparable to that of neuroradiologists.

Understanding Language Reorganization With Neuroimaging: How Language Adapts to Different Focal Lesions and Insights Into Clinical Applications.

When the language-dominant hemisphere is damaged by a focal lesion, the brain may reorganize the language network through functional and structural changes known as adaptive plasticity. Adaptive plasticity is documented for triggers including ischemic, tumoral, and epileptic focal lesions, with effects in clinical practice. Many questions remain regarding language plasticity. Different lesions may induce different patterns of reorganization depending on pathologic features, location in the brain, and timing of onset. Neuroimaging provides insights into language plasticity due to its non-invasiveness, ability to image the whole brain, and large-scale implementation. This review provides an overview of language plasticity on MRI with insights for patient care. First, we describe the structural and functional language network as depicted by neuroimaging. Second, we explore language reorganization triggered by stroke, brain tumors, and epileptic lesions and analyze applications in clinical diagnosis and treatment planning. By comparing different focal lesions, we investigate determinants of language plasticity including lesion location and timing of onset, longitudinal evolution of reorganization, and the relationship between structural and functional changes.

Semisupervised Training of a Brain MRI Tumor Detection Model Using Mined Annotations.

Background Artificial intelligence (AI) applications for cancer imaging conceptually begin with automated tumor detection, which can provide the foundation for downstream AI tasks. However, supervised training requires many image annotations, and performing dedicated post hoc image labeling is burdensome and costly. Purpose To investigate whether clinically generated image annotations can be data mined from the picture archiving and communication system (PACS), automatically curated, and used for semisupervised training of a brain MRI tumor detection model. Materials and Methods In this retrospective study, the cancer center PACS was mined for brain MRI scans acquired between January 2012 and December 2017 and included all annotated axial T1 postcontrast images. Line annotations were converted to boxes, excluding boxes shorter than 1 cm or longer than 7 cm. The resulting boxes were used for supervised training of object detection models using RetinaNet and Mask region-based convolutional neural network (R-CNN) architectures. The best-performing model trained from the mined data set was used to detect unannotated tumors on training images themselves (self-labeling), automatically correcting many of the missing labels. After self-labeling, new models were trained using this expanded data set. Models were scored for precision, recall, and F1 using a held-out test data set comprising 754 manually labeled images from 100 patients (403 intra-axial and 56 extra-axial enhancing tumors). Model F1 scores were compared using bootstrap resampling. Results The PACS query extracted 31 150 line annotations, yielding 11 880 boxes that met inclusion criteria. This mined data set was used to train models, yielding F1 scores of 0.886 for RetinaNet and 0.908 for Mask R-CNN. Self-labeling added 18 562 training boxes, improving model F1 scores to 0.935 (P < .001) and 0.954 (P < .001), respectively. Conclusion The application of semisupervised learning to mined image annotations significantly improved tumor detection performance, achieving an excellent F1 score of 0.954. This development pipeline can be extended for other imaging modalities, repurposing unused data silos to potentially enable automated tumor detection across radiologic modalities. © RSNA, 2022 Online supplemental material is available for this article.