Vascular architecture mapping for early detection of glioblastoma recurrence.

OBJECTIVE: Treatment failure and inevitable tumor recurrence are the main reasons for the poor prognosis of glioblastoma (GB). Gross-total resection at repeat craniotomy for GB recurrence improves patient overall survival but requires early and reliable detection. It is known, however, that even advanced MRI approaches have limited diagnostic performance for distinguishing tumor progression from pseudoprogression. The novel MRI technique of vascular architectural mapping (VAM) provides deeper insight into tumor microvascularity and neovascularization. In this study the authors evaluated the usefulness of VAM for the monitoring of GB patients and quantitatively analyzed the features of neovascularization of early- and progressed-stage GB recurrence. METHODS: In total, a group of 115 GB patients who received overall 374 follow-up MRI examinations after standard treatment were retrospectively evaluated in this study. The clinical routine MRI (cMRI) protocol at 3 Tesla was extended with the authors' experimental VAM approach, requiring 2 minutes of extra time for data acquisition. Custom-made MATLAB software was used for calculation of imaging biomarker maps of macrovascular perfusion from perfusion cMRI as well as of microvascular perfusion and architecture from VAM data. Additionally, cMRI data were analyzed by two board-certified radiologists in consensus. Statistical procedures included receiver operating characteristic (ROC) analysis to determine diagnostic performances for GB recurrence detection. RESULTS: Overall, cMRI showed GB recurrence in 89 patients, and in 28 of these patients recurrence was detected earlier with VAM data, by 1 (20 patients) or 2 (8 patients) follow-up examinations, than with cMRI data. The mean time difference between recurrence detection with VAM and cMRI data was 147 days. During this time period the mean tumor volume increased significantly (p < 0.001) from 9.7 to 26.8 cm3. Quantitative analysis of imaging biomarkers demonstrated microvascular but no macrovascular hyperperfusion in early GB recurrence. Therefore, ROC analysis revealed superior diagnostic performance for VAM compared with cMRI. CONCLUSIONS: This study demonstrated that the targeted assessment of microvascular features using the VAM technique provided valuable information about early neovascularization activity in recurrent GB that is complementary to perfusion cMRI and may be helpful for earlier and more precise monitoring of patients suffering from GB. This VAM approach is compatible with existing cMRI protocols. Prospective clinical trials are necessary to investigate the clinical usefulness and potential benefit of increased overall survival with the use of VAM in patients with recurrent GB.

Machine learning applications for the differentiation of primary central nervous system lymphoma from glioblastoma on imaging: a systematic review and meta-analysis.

OBJECTIVEGlioblastoma (GBM) and primary central nervous system lymphoma (PCNSL) are common intracranial pathologies encountered by neurosurgeons. They often may have similar radiological findings, making diagnosis difficult without surgical biopsy; however, management is quite different between these two entities. Recently, predictive analytics, including machine learning (ML), have garnered attention for their potential to aid in the diagnostic assessment of a variety of pathologies. Several ML algorithms have recently been designed to differentiate GBM from PCNSL radiologically with a high sensitivity and specificity. The objective of this systematic review and meta-analysis was to evaluate the implementation of ML algorithms in differentiating GBM and PCNSL.METHODSThe authors performed a systematic review of the literature using PubMed in accordance with PRISMA guidelines to select and evaluate studies that included themes of ML and brain tumors. These studies were further narrowed down to focus on works published between January 2008 and May 2018 addressing the use of ML in training models to distinguish between GBM and PCNSL on radiological imaging. Outcomes assessed were test characteristics such as accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC).RESULTSEight studies were identified addressing use of ML in training classifiers to distinguish between GBM and PCNSL on radiological imaging. ML performed well with the lowest reported AUC being 0.878. In studies in which ML was directly compared with radiologists, ML performed better than or as well as the radiologists. However, when ML was applied to an external data set, it performed more poorly.CONCLUSIONSFew studies have applied ML to solve the problem of differentiating GBM from PCNSL using imaging alone. Of the currently published studies, ML algorithms have demonstrated promising results and certainly have the potential to aid radiologists with difficult cases, which could expedite the neurosurgical decision-making process. It is likely that ML algorithms will help to optimize neurosurgical patient outcomes as well as the cost-effectiveness of neurosurgical care if the problem of overfitting can be overcome.

Prospective evaluation of the utility of intraoperative confocal laser endomicroscopy in patients with brain neoplasms using fluorescein sodium: experience with 74 cases.

OBJECTIVE: This study evaluated the utility, specificity, and sensitivity of intraoperative confocal laser endomicroscopy (CLE) to provide diagnostic information during resection of human brain tumors. METHODS: CLE imaging was used in the resection of intracranial neoplasms in 74 consecutive patients (31 male; mean age 47.5 years; sequential 10-month study period). Intraoperative in vivo and ex vivo CLE was performed after intravenous injection of fluorescein sodium (FNa). Tissue samples from CLE imaging-matched areas were acquired for comparison with routine histological analysis (frozen and permanent sections). CLE images were classified as diagnostic or nondiagnostic. The specificities and sensitivities of CLE and frozen sections for gliomas and meningiomas were calculated using permanent histological sections as the standard. RESULTS: CLE images were obtained for each patient. The mean duration of intraoperative CLE system use was 15.7 minutes (range 3-73 minutes). A total of 20,734 CLE images were correlated with 267 biopsy specimens (mean number of images/biopsy location, in vivo 84, ex vivo 70). CLE images were diagnostic for 45.98% in vivo and 52.97% ex vivo specimens. After initiation of CLE, an average of 14 in vivo images and 7 ex vivo images were acquired before identification of a first diagnostic image. CLE specificity and sensitivity were, respectively, 94% and 91% for gliomas and 93% and 97% for meningiomas. CONCLUSIONS: CLE with FNa provided intraoperative histological information during brain tumor removal. Specificities and sensitivities of CLE for gliomas and meningiomas were comparable to those for frozen sections. These data suggest that CLE could allow the interactive identification of tumor areas, substantially improving intraoperative decisions during the resection of brain tumors.

T2 mapping of lumbosacral nerves in patients suffering from unilateral radicular pain due to degenerative disc disease.

OBJECTIVELumbosacral radicular syndrome (LRS) is a very common condition, often requiring diagnostic imaging with the aim of elucidating a structural cause when symptoms are longer lasting. However, findings on conventional anatomical MRI do not necessarily correlate with clinical symptoms, and it is primarily performed for the qualitative evaluation of surrounding compressive structures, such as herniated discs, instead of to evaluate the nerves directly. The present study investigated the performance of quantitative imaging by using magnetic resonance neurography (MRN) in patients with LRS.METHODSEighteen patients (55.6% males, mean age 64.4 ± 10.2 years), with strict unilateral LRS matching at least one dermatome and suspected disc herniation, underwent high-resolution 3-T MRN using T2 mapping. On T2 maps, the presumably affected and contralateral unaffected nerves were identified; subsequent regions of interest (ROIs) were placed at preganglionic, ganglionic, and postganglionic sites; and T2 values were extracted. Patients then underwent an epidural steroid injection (ESI) with local anesthetic agents at the site of suspected nerve affection. T2 values of the affected nerves were compared against the contralateral nerves. Furthermore, receiver operating characteristics were calculated based on the measured T2 values and the responsiveness to ESI.RESULTSThe mean T2 value was 77.3 ± 1.9 msec for affected nerves and 74.8 ± 1.4 msec for contralateral nerves (p < 0.0001). In relation to ESI performed at the site of suspected nerve affection, MRN with T2 mapping had a sensitivity/specificity of 76.9%/60.0% and a positive/negative predictive value of 83.3%/50.0%. Signal alterations in affected nerves according to qualitative visual inspection were present in only 22.2% of patients.CONCLUSIONSAs one of the first of its kind, this study revealed elevated T2 values in patients suffering from LRS. T2 values of lumbosacral nerves might be used as more objective parameters to directly detect nerve affection in such patients.

Interhemispheric connectivity revealed by diffusion tensor imaging fiber tracking derived from navigated transcranial magnetic stimulation maps as a sign of language function at risk in patients with brain tumors.

OBJECTIVE Resection of brain tumors in language-eloquent areas entails the risk of postoperative aphasia. It has been demonstrated via navigated transcranial magnetic stimulation (nTMS) that language function can partially shift to the unaffected hemisphere due to tumor-induced plasticity. Therefore, this study was designed to evaluate whether interhemispheric connectivity (IC) detected by nTMS-based diffusion tensor imaging-fiber tracking (DTI-FT) can be used to predict surgery-related aphasia in patients with brain tumors. METHODS Thirty-eight patients with left-sided perisylvian brain lesions underwent cortical language mapping of both hemispheres by nTMS prior to awake surgery. Then, nTMS-based DTI-FT was conducted with a fractional anisotropy (FA) of 0.01 and 0.2 to visualize nTMS-based IC. Receiver operating characteristics were calculated for the prediction of a postoperative (irrespective of the preoperative state) and a new surgery-related aphasia by the presence of detectable IC. RESULTS Language mapping by nTMS was possible in all patients. Seventeen patients (44.7%) suffered from surgery-related worsening of language performance (transient aphasia according to 3-month follow-up in 16 subjects [42.1%]; new permanent aphasia according to 3-month follow-up in 1 patient [2.6%]). Regarding the correlation of aphasia to nTMS-based IC, statistically significant differences were revealed for both evaluated FA values. However, better results were observed for tractography with an FA of 0.2, which led to a specificity of 93% (postoperative aphasia) and 90% (surgery-related aphasia). For postoperative aphasia, the corresponding OR was 0.1282 (95% CI 0.0143-1.1520), and for surgery-related aphasia the OR was 0.1184 (95% CI 0.0208-0.6754). CONCLUSIONS According to these results, IC detected by preoperative nTMS-based DTI-FT might be regarded as a risk factor for surgery-related aphasia, with a specificity of up to 93%. However, because the majority of enrolled patients suffered from transient aphasia postoperatively, it has to be evaluated whether this approach distinctly leads to similar results among patients with permanent language deficits. Despite this restriction, this approach might contribute to individualized patient consultation prior to tumor resection in clinical practice.