Cellular Density in Adult Glioma, Estimated with MR Imaging Data and a Machine Learning Algorithm, Has Prognostic Power Approaching World Health Organization Histologic Grading in a Cohort of 1181 Patients.

BACKGROUND AND PURPOSE: Recent advances in machine learning have enabled image-based prediction of local tissue pathology in gliomas, but the clinical usefulness of these predictions is unknown. We aimed to evaluate the prognostic ability of imaging-based estimates of cellular density for patients with gliomas, with comparison to the gold standard reference of World Health Organization grading. MATERIALS AND METHODS: Data from 1181 (207 grade II, 246 grade III, 728 grade IV) previously untreated patients with gliomas from a single institution were analyzed. A pretrained random forest model estimated voxelwise tumor cellularity using MR imaging data. Maximum cellular density was correlated with the World Health Organization grade and actual survival, correcting for covariates of age and performance status. RESULTS: A maximum estimated cellular density of >7681 nuclei/mm2 was associated with a worse prognosis and a univariate hazard ratio of 4.21 (P < .001); the multivariate hazard ratio after adjusting for covariates of age and performance status was 2.91 (P < .001). The concordance index between maximum cellular density (adjusted for covariates) and survival was 0.734. The hazard ratio for a high World Health Organization grade (IV) was 7.57 univariate (P < .001) and 5.25 multivariate (P < .001). The concordance index for World Health Organization grading (adjusted for covariates) was 0.761. The maximum cellular density was an independent predictor of overall survival, and a Cox model using World Health Organization grade, maximum cellular density, age, and Karnofsky performance status had a higher concordance (C = 0.764; range 0.748-0.781) than the component predictors. CONCLUSIONS: Image-based estimation of glioma cellularity is a promising biomarker for predicting survival, approaching the prognostic power of World Health Organization grading, with added values of early availability, low risk, and low cost.

Estimating Local Cellular Density in Glioma Using MR Imaging Data.

BACKGROUND AND PURPOSE: Increased cellular density is a hallmark of gliomas, both in the bulk of the tumor and in areas of tumor infiltration into surrounding brain. Altered cellular density causes altered imaging findings, but the degree to which cellular density can be quantitatively estimated from imaging is unknown. The purpose of this study was to discover the best MR imaging and processing techniques to make quantitative and spatially specific estimates of cellular density. MATERIALS AND METHODS: We collected stereotactic biopsies in a prospective imaging clinical trial targeting untreated patients with gliomas at our institution undergoing their first resection. The data included preoperative MR imaging with conventional anatomic, diffusion, perfusion, and permeability sequences and quantitative histopathology on biopsy samples. We then used multiple machine learning methodologies to estimate cellular density using local intensity information from the MR images and quantitative cellular density measurements at the biopsy coordinates as the criterion standard. RESULTS: The random forest methodology estimated cellular density with R 2 = 0.59 between predicted and observed values using 4 input imaging sequences chosen from our full set of imaging data (T2, fractional anisotropy, CBF, and area under the curve from permeability imaging). Limiting input to conventional MR images (T1 pre- and postcontrast, T2, and FLAIR) yielded slightly degraded performance (R2 = 0.52). Outputs were also reported as graphic maps. CONCLUSIONS: Cellular density can be estimated with moderate-to-strong correlations using MR imaging inputs. The random forest machine learning model provided the best estimates. These spatially specific estimates of cellular density will likely be useful in guiding both diagnosis and treatment.

Distinguishing Recurrent Thyroid Cancer from Residual Nonmalignant Thyroid Tissue Using Multiphasic Multidetector CT.

BACKGROUND AND PURPOSE: During thyroidectomy incomplete resection of the thyroid gland may occur. This complicates the imaging surveillance of these patients as residual thyroid needs to be distinguished from local recurrence. Therefore, the purpose of this study was to determine if multiphasic multi-detector computed tomography (4D-MDCT) can differentiate residual nonmalignant thyroid tissue and recurrent thyroid carcinoma after thyroidectomy. MATERIALS AND METHODS: In this retrospective study, Hounsfield unit values on multiphasic multidetector CT in precontrast, arterial (25 seconds), venous (55 seconds), and delayed (85 seconds) phases were compared in 29 lesions of recurrent thyroid cancer, 29 with normal thyroid, and 29 with diseased thyroid (thyroiditis/multinodular thyroid). The comparison of Hounsfield unit values among lesion types by phase was performed using ANOVA. The performance of Hounsfield unit values to predict recurrence was evaluated by logistic regression and receiver operating characteristic analysis. RESULTS: All 3 tissue types had near-parallel enhancement characteristics, with a wash-in-washout pattern. Statistically different Hounsfield unit density was noted between the recurrence (lowest Hounsfield unit), diseased (intermediate Hounsfield unit), and normal (highest Hounsfield unit) thyroid groups throughout all 4 phases (P < .001 for each group and in each phase). Dichotomized recurrence-versus-diseased/normal thyroid tissue with univariate logistic regression analysis demonstrated that the area under the receiver operating characteristic curve for differentiating benign from malignant thyroid for the various phases of enhancement was greatest in the precontrast phase at 0.983 (95% CI, 0.954-1), with a cutoff value of ≤62 (sensitivity/specificity, 0.966/0.983) followed by the arterial phase. CONCLUSIONS: Recurrent thyroid carcinoma can be distinguished from residual nonmalignant thyroid tissue using multiphasic multidetector CT with high accuracy. The maximum information for discrimination is in the precontrast images, then the arterial phase. An optimal clinical protocol could be built from any number of phases but should include a precontrast phase.

Imaging-Based Algorithm for the Local Grading of Glioma.

BACKGROUND AND PURPOSE: Gliomas are highly heterogeneous tumors, and optimal treatment depends on identifying and locating the highest grade disease present. Imaging techniques for doing so are generally not validated against the histopathologic criterion standard. The purpose of this work was to estimate the local glioma grade using a machine learning model trained on preoperative image data and spatially specific tumor samples. The value of imaging in patients with brain tumor can be enhanced if pathologic data can be estimated from imaging input using predictive models. MATERIALS AND METHODS: Patients with gliomas were enrolled in a prospective clinical imaging trial between 2013 and 2016. MR imaging was performed with anatomic, diffusion, permeability, and perfusion sequences, followed by image-guided stereotactic biopsy before resection. An imaging description was developed for each biopsy, and multiclass machine learning models were built to predict the World Health Organization grade. Models were assessed on classification accuracy, Cohen κ, precision, and recall. RESULTS: Twenty-three patients (with 7/9/7 grade II/III/IV gliomas) had analyzable imaging-pathologic pairs, yielding 52 biopsy sites. The random forest method was the best algorithm tested. Tumor grade was predicted at 96% accuracy (κ = 0.93) using 4 inputs (T2, ADC, CBV, and transfer constant from dynamic contrast-enhanced imaging). By means of the conventional imaging only, the overall accuracy decreased (89% overall, κ = 0.79) and 43% of high-grade samples were misclassified as lower-grade disease. CONCLUSIONS: We found that local pathologic grade can be predicted with a high accuracy using clinical imaging data. Advanced imaging data improved this accuracy, adding value to conventional imaging. Confirmatory imaging trials are justified.

Solitary Parathyroid Adenoma Localization in Technetium Tc99m Sestamibi SPECT and Multiphase Multidetector 4D CT.

BACKGROUND AND PURPOSE: Minimally invasive parathyroid surgery relies critically on image guidance, but data comparing the efficacy of various imaging modalities are scarce. Our aim was to perform a blinded comparison of the localizing capability of technetium Tc99m sestamibi SPECT, multiphase multidetector 4D CT, and the combination of these 2 modalities (technetium Tc99m sestamibi SPECT + multiphase multidetector 4D CT). MATERIALS AND METHODS: We reviewed the records of 31 (6 men, 25 women; median age, 56 years) consecutive patients diagnosed with biochemically confirmed primary hyperparathyroidism between November 2009 and March 2010 who underwent preoperative technetium Tc99m sestamibi SPECT and multiphase multidetector 4D CT performed on the same scanner with pathologic confirmation by resection of a single parathyroid adenoma. Accuracy was determined separately for localization to the correct side and quadrant using surgical localization as the standard of reference. RESULTS: Surgical resection identified 14 left and 17 right parathyroid adenomas and 2 left inferior, 12 left superior, 11 right inferior, and 6 right superior parathyroid adenomas. For left/right localization, technetium Tc99m sestamibi SPECT achieved an accuracy of 93.5% (29 of 31), multiphase multidetector 4D CT achieved 96.8% accuracy (30 of 31), and technetium Tc99m sestamibi SPECT + multiphase multidetector 4D CT achieved 96.8% accuracy (30 of 31). For quadrant localization, technetium Tc99m sestamibi SPECT accuracy was 67.7% (21 of 31), multiphase multidetector 4D CT accuracy was 87.1% (27 of 31), and technetium Tc99m sestamibi SPECT + multiphase multidetector 4D CT accuracy was 93.5% (29 of 31). Reader diagnostic confidence was consistently ranked lowest for technetium Tc99m sestamibi SPECT and highest for technetium Tc99m sestamibi SPECT + multiphase multidetector 4D CT. CONCLUSIONS: For left/right localization of parathyroid adenomas, all modalities performed equivalently. For quadrant localization, technetium Tc99m sestamibi SPECT + multiphase multidetector 4D CT is superior to technetium Tc99m sestamibi SPECT.

Performance Assessment for Brain MR Imaging Registration Methods.

BACKGROUND AND PURPOSE: Clinical brain MR imaging registration algorithms are often made available by commercial vendors without figures of merit. The purpose of this study was to suggest a rational performance comparison methodology for these products. MATERIALS AND METHODS: Twenty patients were imaged on clinical 3T scanners by using 4 sequences: T2-weighted, FLAIR, susceptibility-weighted angiography, and T1 postcontrast. Fiducial landmark sites (n = 1175) were specified throughout these image volumes to define identical anatomic locations across sequences. Multiple registration algorithms were applied by using the T2 sequence as a fixed reference. Euclidean error was calculated before and after each registration and compared with a criterion standard landmark registration. The Euclidean effectiveness ratio is the fraction of Euclidean error remaining after registration, and the statistical effectiveness ratio is similar, but accounts for dispersion and noise. RESULTS: Before registration, error values for FLAIR, susceptibility-weighted angiography, and T1 postcontrast were 2.07 ± 0.55 mm, 2.63 ± 0.62 mm, and 3.65 ± 2.00 mm, respectively. Postregistration, the best error values for FLAIR, susceptibility-weighted angiography, and T1 postcontrast were 1.55 ± 0.46 mm, 1.34 ± 0.23 mm, and 1.06 ± 0.16 mm, with Euclidean effectiveness ratio values of 0.493, 0.181, and 0.096 and statistical effectiveness ratio values of 0.573, 0.352, and 0.929 for rigid mutual information, affine mutual information, and a commercial GE registration, respectively. CONCLUSIONS: We demonstrate a method for comparing the performance of registration algorithms and suggest the Euclidean error, Euclidean effectiveness ratio, and statistical effectiveness ratio as performance metrics for clinical registration algorithms. These figures of merit allow registration algorithms to be rationally compared.

Imaging and clinical features of an intra-axial brain stem schwannoma.

A 68-year-old man presented with a highly symptomatic brain stem tumor originally thought to be a brain stem glioma. Intraoperative MR imaging guidance was used to resect the tumor, and real-time evoked potentials improved during surgery. Pathology findings unexpectedly indicated that the tumor was an intra-axial brain stem schwannoma, a condition reported, to our knowledge, only 6 times previously in the literature. The patient made an excellent recovery with reversal of his symptoms.

Multidetector CT-Guided Lumbar Puncture in Patients with Cancer.

SUMMARY: Lumbar puncture can be performed for therapeutic purposes, to instill intrathecal chemotherapy for leptomeningeal cancer treatment or prophylaxis. This technique is generally performed blindly or under fluoroscopic guidance. However, in certain situations, lumbar puncture using multidetector CT (MDCT)-guided imaging may be beneficial, when other options have been exhausted or depending on the requirements of the performing radiologist's institution. The purpose of this article is to describe the technique and to evaluate outcomes of MDCT-guided lumber puncture for diagnostic and therapeutic purposes in patients with cancer. We conclude that MDCT-guided lumbar puncture is an effective and safe guiding modality for thecal sac access in patients with cancer, particularly where other methods of intrathecal access have failed.

Focal cement accumulation in lytic metastatic lesions: a new sign in vertebroplasty?

SUMMARY: We describe a unique imaging appearance of focal cement accumulation inside lytic metastatic lesions in four patients with pathologic vertebral fractures undergoing vertebroplasty. This appearance differs from the normal, trabecular appearance of cement distribution in osteoporotic vertebrae. Two patients, in whom the lytic metastasis was completely filled with cement, had a complete pain response, and two patients, in whom the lytic metastasis could only be partially filled, had an incomplete response. Focal cement accumulation in a lytic lesion may be a predictor of favorable patient outcome in patients with lytic metastasis.

Quantitation of HSV mass distribution in a rodent brain tumor model.

A number of different viral vectors have been used for gene therapy of tumors, with many more under construction, ultimately designed to improve tumor targeting and transduction efficiency. It has become apparent that insufficient viral delivery can be a key limitation to treatment efficacy. We have studied the in vivo mass distribution of a herpes simplex virus type 1 (HSV) vector, hrR3, by radiolabeling it with 111In-oxine. The virus was administered to intracerebral 9L glioma bearing Fisher (F-344) rats by intracarotid and intratumoral injection. The blood half-life of the virus was 1 min (fast component, 10% contribution) and 180 min (slow component, 90% contribution). Approximately 20% of activity had been excreted by 24 h. With intracarotid injection, the total amount of virus that accumulated in tumor was 0.10+/-0.07% of the injected dose (ID)/g at 1 h and 0.19+/-0.01% ID/g at 24 h. By comparison, co-injection of RMP-7, a synthetic bradykinin analog, with the virus, resulted in slightly increased tumor delivery of 0.17+/-0.10% ID/g (P 0.05) at 1 h. The 1 h organ distribution after intra-arterial injection (%ID/organ) was as follows: liver 273+/-2.86%, lung 2.10+/-0.68% and kidney 1.78+/-1.60% with lesser amounts in other organs. When virus was injected directly into the tumor, 71% of virus remained in tumor at 24 h (590+/-212 %ID/g, consistent with the small tumor mass containing most of the virus) with the following distribution regions: tumor > border zone > normal brain (99:40: 1). These studies are the first quantitative mass distribution studies of HSV vectors in an experimental brain tumor model. Localization and quantitation of viral accumulation in vivo will enable detailed analysis of viral and organ interactions critical for advancing the therapeutic use of vectors.

Mapping the in vivo distribution of herpes simplex virions.

We describe a method for labeling enveloped viral particles with a radiotracer, indium-111, allowing labeled viruses to be traced in vivo by nuclear imaging. After initial optimization experiments, a labeling efficiency of 83% (incorporation yield) was achieved for herpes simplex virus (HSV), resulting in a specific activity of 30 microCi/10(9) PFU. The labeling procedure did not significantly reduce the infectivity of the labeled virus and the virus did not release any significant amounts of the radionuclide within 12 hr after labeling. Sequential imaging of animals after intravenous administration of the labeled virus showed fast accumulation in the liver and redistribution from the blood pool (immediately after injection) to liver and spleen (12-24 hr after injection). At 12 hr after injection 7% of the virus-associated (111)In had been eliminated from the body and the remaining organ distribution of the virus was as follows: spleen 2.87 +/- 0.54% ID/g; liver, 2.60 +/- 0.51% ID/g; kidney, 0.98 +/- 0.31% ID/g; lung, 0.57 +/- 0.10% ID/g; [corrected] and lower amounts in other organs. Our results indicate that the described method allows qualitative and quantitative assessment of viral biodistribution in vivo by nuclear imaging.