Michel Bilello

GLISTR: Glioma Image Segmentation and Registration

We present a generative approach for simultaneously registering a probabilistic atlas of a healthy population to brain magnetic resonance (MR) scans showing glioma and segmenting the scans into tumor as well as healthy tissue labels. The proposed method is based on the expectation maximization (EM) algorithm that incorporates a glioma growth model for atlas seeding, a process which modifies the original atlas into one with tumor and edema adapted to best match a given set of patients images. The modified atlas is registered into the patient space and utilized for estimating the posterior probabilities of various tissue labels. EM iteratively refines the estimates of the posterior probabilities of tissue labels, the deformation field and the tumor growth model parameters. Hence, in addition to segmentation, the proposed method results in atlas registration and a low-dimensional description of the patient scans through estimation of tumor model parameters. We validate the method by automatically segmenting 10 MR scans and comparing the results to those produced by clinical experts and two state-of-the-art methods. The resulting segmentations of tumor and edema outperform the results of the reference methods, and achieve a similar accuracy from a second human rater. We additionally apply the method to 122 patients scans and report the estimated tumor model parameters and their relations with segmentation and registration results. Based on the results from this patient population, we construct a statistical atlas of the glioma by inverting the estimated deformation fields to warp the tumor segmentations of patients scans into a common space.

Stroke After Aortic Valve Surgery Results From a Prospective Cohort

Background— The incidence and impact of clinical stroke and silent radiographic cerebral infarction complicating open surgical aortic valve replacement (AVR) are poorly characterized. Methods and Results— We performed a prospective cohort study of subjects ≥65 years of age who were undergoing AVR for calcific aortic stenosis. Subjects were evaluated by neurologists preoperatively and postoperatively and underwent postoperative magnetic resonance imaging. Over a 4-year period, 196 subjects were enrolled at 2 sites (mean age, 75.8±6.2 years; 36% women; 6% nonwhite). Clinical strokes were detected in 17%, transient ischemic attack in 2%, and in-hospital mortality was 5%. The frequency of stroke in the Society for Thoracic Surgery database in this cohort was 7%. Most strokes were mild; the median National Institutes of Health Stroke Scale was 3 (interquartile range, 1–9). Clinical stroke was associated with increased length of stay (median, 12 versus 10 days; P=0.02). Moderate or severe stroke (National Institutes of Health Stroke Scale ≥10) occurred in 8 (4%) and was strongly associated with in-hospital mortality (38% versus 4%; P=0.005). Of the 109 stroke-free subjects with postoperative magnetic resonance imaging, silent infarct was identified in 59 (54%). Silent infarct was not associated with in-hospital mortality or increased length of stay. Conclusions— Clinical stroke after AVR was more common than reported previously, more than double for this same cohort in the Society for Thoracic Surgery database, and silent cerebral infarctions were detected in more than half of the patients undergoing AVR. Clinical stroke complicating AVR is associated with increased length of stay and mortality.Background— The incidence and impact of clinical stroke and silent radiographic cerebral infarction complicating open surgical aortic valve replacement (AVR) are poorly characterized. Methods and Results— We performed a prospective cohort study of subjects ≥65 years of age who were undergoing AVR for calcific aortic stenosis. Subjects were evaluated by neurologists preoperatively and postoperatively and underwent postoperative magnetic resonance imaging. Over a 4-year period, 196 subjects were enrolled at 2 sites (mean age, 75.8±6.2 years; 36% women; 6% nonwhite). Clinical strokes were detected in 17%, transient ischemic attack in 2%, and in-hospital mortality was 5%. The frequency of stroke in the Society for Thoracic Surgery database in this cohort was 7%. Most strokes were mild; the median National Institutes of Health Stroke Scale was 3 (interquartile range, 1–9). Clinical stroke was associated with increased length of stay (median, 12 versus 10 days; P =0.02). Moderate or severe stroke (National Institutes of Health Stroke Scale ≥10) occurred in 8 (4%) and was strongly associated with in-hospital mortality (38% versus 4%; P =0.005). Of the 109 stroke-free subjects with postoperative magnetic resonance imaging, silent infarct was identified in 59 (54%). Silent infarct was not associated with in-hospital mortality or increased length of stay. Conclusions— Clinical stroke after AVR was more common than reported previously, more than double for this same cohort in the Society for Thoracic Surgery database, and silent cerebral infarctions were detected in more than half of the patients undergoing AVR. Clinical stroke complicating AVR is associated with increased length of stay and mortality. # CLINICAL PERSPECTIVE {#article-title-47}

Imaging patterns predict patient survival and molecular subtype in glioblastoma via machine learning techniques.

BACKGROUND MRI characteristics of brain gliomas have been used to predict clinical outcome and molecular tumor characteristics. However, previously reported imaging biomarkers have not been sufficiently accurate or reproducible to enter routine clinical practice and often rely on relatively simple MRI measures. The current study leverages advanced image analysis and machine learning algorithms to identify complex and reproducible imaging patterns predictive of overall survival and molecular subtype in glioblastoma (GB). METHODS One hundred five patients with GB were first used to extract approximately 60 diverse features from preoperative multiparametric MRIs. These imaging features were used by a machine learning algorithm to derive imaging predictors of patient survival and molecular subtype. Cross-validation ensured generalizability of these predictors to new patients. Subsequently, the predictors were evaluated in a prospective cohort of 29 new patients. RESULTS Survival curves yielded a hazard ratio of 10.64 for predicted long versus short survivors. The overall, 3-way (long/medium/short survival) accuracy in the prospective cohort approached 80%. Classification of patients into the 4 molecular subtypes of GB achieved 76% accuracy. CONCLUSIONS By employing machine learning techniques, we were able to demonstrate that imaging patterns are highly predictive of patient survival. Additionally, we found that GB subtypes have distinctive imaging phenotypes. These results reveal that when imaging markers related to infiltration, cell density, microvascularity, and blood-brain barrier compromise are integrated via advanced pattern analysis methods, they form very accurate predictive biomarkers. These predictive markers used solely preoperative images, hence they can significantly augment diagnosis and treatment of GB patients.

Joint segmentation and deformable registration of brain scans guided by a tumor growth model

This paper presents an approach for joint segmentation and deformable registration of brain scans of glioma patients to a normal atlas. The proposed method is based on the Expectation Maximization (EM) algorithm that incorporates a glioma growth model for atlas seeding, a process which modifies the normal atlas into one with a tumor and edema. The modified atlas is registered into the patient space and utilized for the posterior probability estimation of various tissue labels. EM iteratively refines the estimates of the registration parameters, the posterior probabilities of tissue labels and the tumor growth model parameters. We have applied this approach to 10 glioma scans acquired with four Magnetic Resonance (MR) modalities (T1, T1-CE, T2 and FLAIR) and validated the result by comparing them to manual segmentations by clinical experts. The resulting segmentations look promising and quantitatively match well with the expert provided ground truth.

Direct visualization of short transverse relaxation time component (ViSTa)

White matter of the brain has been demonstrated to have multiple relaxation components. Among them, the short transverse relaxation time component (T2<40 ms; T2⁎<25 ms at 3 T) has been suggested to originate from myelin water whereas long transverse relaxation time components have been associated with axonal and/or interstitial water. In myelin water imaging, T2 or T2⁎ signal decay is measured to estimate myelin water fraction based on T2 or T2⁎ differences among the water components. This method has been demonstrated to be sensitive to demyelination in the brain but suffers from low SNR and image artifacts originating from ill-conditioned multi-exponential fitting. In this study, a novel approach that selectively acquires short transverse relaxation time signal is proposed. The method utilizes a double inversion RF pair to suppress a range of long T1 signal. This suppression leaves short T2⁎ signal, which has been suggested to have short T1, as the primary source of the image. The experimental results confirm that after suppression of long T1 signals, the image is dominated by short T2⁎ in the range of myelin water, allowing us to directly visualize the short transverse relaxation time component in the brain. Compared to conventional myelin water imaging, this new method of direct visualization of short relaxation time component (ViSTa) provides high quality images. When applied to multiple sclerosis patients, chronic lesions show significantly reduced signal intensity in ViSTa images suggesting sensitivity to demyelination.

Pattern Analysis of Dynamic Susceptibility Contrast-enhanced MR Imaging Demonstrates Peritumoral Tissue Heterogeneity

PURPOSE To augment the analysis of dynamic susceptibility contrast material-enhanced magnetic resonance (MR) images to uncover unique tissue characteristics that could potentially facilitate treatment planning through a better understanding of the peritumoral region in patients with glioblastoma. MATERIALS AND METHODS Institutional review board approval was obtained for this study, with waiver of informed consent for retrospective review of medical records. Dynamic susceptibility contrast-enhanced MR imaging data were obtained for 79 patients, and principal component analysis was applied to the perfusion signal intensity. The first six principal components were sufficient to characterize more than 99% of variance in the temporal dynamics of blood perfusion in all regions of interest. The principal components were subsequently used in conjunction with a support vector machine classifier to create a map of heterogeneity within the peritumoral region, and the variance of this map served as the heterogeneity score. RESULTS The calculated principal components allowed near-perfect separability of tissue that was likely highly infiltrated with tumor and tissue that was unlikely infiltrated with tumor. The heterogeneity map created by using the principal components showed a clear relationship between voxels judged by the support vector machine to be highly infiltrated and subsequent recurrence. The results demonstrated a significant correlation (r = 0.46, P < .0001) between the heterogeneity score and patient survival. The hazard ratio was 2.23 (95% confidence interval: 1.4, 3.6; P < .01) between patients with high and low heterogeneity scores on the basis of the median heterogeneity score. CONCLUSION Analysis of dynamic susceptibility contrast-enhanced MR imaging data by using principal component analysis can help identify imaging variables that can be subsequently used to evaluate the peritumoral region in glioblastoma. These variables are potentially indicative of tumor infiltration and may become useful tools in guiding therapy, as well as individualized prognostication.

Correlating Cognitive Decline with White Matter Lesion and Brain Atrophy Magnetic Resonance Imaging Measurements in Alzheimer's Disease.

BACKGROUND Vascular risk factors are increasingly recognized as risks factors for Alzheimers disease (AD) and early conversion from mild cognitive impairment (MCI) to dementia. While neuroimaging research in AD has focused on brain atrophy, metabolic function, or amyloid deposition, little attention has been paid to the effect of cerebrovascular disease to cognitive decline. OBJECTIVE To investigate the correlation of brain atrophy and white matter lesions with cognitive decline in AD, MCI, and control subjects. METHODS Patients with AD and MCI, and healthy subjects were included in this study. Subjects had a baseline MRI scan, and baseline and follow-up neuropsychological battery (CERAD). Regional volumes were measured, and white matter lesion segmentation was performed. Correlations between rate of CERAD score decline and white matter lesion load and brain structure volume were evaluated. In addition, voxel-based correlations between baseline CERAD scores and atrophy and white matter lesion measures were computed. RESULTS CERAD rate of decline was most significantly associated with lesion loads located in the fornices. Several temporal lobe ROI volumes were significantly associated with CERAD decline. Voxel-based analysis demonstrated strong correlation between baseline CERAD scores and atrophy measures in the anterior temporal lobes. Correlation of baseline CERAD scores with white matter lesion volumes achieved significance in multilobar subcortical white matter. CONCLUSION Both baseline and declines in CERAD scores correlate with white matter lesion load and gray matter atrophy. Results of this study highlight the dominant effect of volume loss, and underscore the importance of small vessel disease as a contributor to cognitive decline in the elderly.

Imaging Surrogates of Infiltration Obtained Via Multiparametric Imaging Pattern Analysis Predict Subsequent Location of Recurrence of Glioblastoma.

BACKGROUND Glioblastoma is an aggressive and highly infiltrative brain cancer. Standard surgical resection is guided by enhancement on postcontrast T1-weighted (T1) magnetic resonance imaging, which is insufficient for delineating surrounding infiltrating tumor. OBJECTIVE To develop imaging biomarkers that delineate areas of tumor infiltration and predict early recurrence in peritumoral tissue. Such markers would enable intensive, yet targeted, surgery and radiotherapy, thereby potentially delaying recurrence and prolonging survival. METHODS Preoperative multiparametric magnetic resonance images (T1, T1-gadolinium, T2-weighted, T2-weighted fluid-attenuated inversion recovery, diffusion tensor imaging, and dynamic susceptibility contrast-enhanced magnetic resonance images) from 31 patients were combined using machine learning methods, thereby creating predictive spatial maps of infiltrated peritumoral tissue. Cross-validation was used in the retrospective cohort to achieve generalizable biomarkers. Subsequently, the imaging signatures learned from the retrospective study were used in a replication cohort of 34 new patients. Spatial maps representing the likelihood of tumor infiltration and future early recurrence were compared with regions of recurrence on postresection follow-up studies with pathology confirmation. RESULTS This technique produced predictions of early recurrence with a mean area under the curve of 0.84, sensitivity of 91%, specificity of 93%, and odds ratio estimates of 9.29 (99% confidence interval: 8.95-9.65) for tissue predicted to be heavily infiltrated in the replication study. Regions of tumor recurrence were found to have subtle, yet fairly distinctive multiparametric imaging signatures when analyzed quantitatively by pattern analysis and machine learning. CONCLUSION Visually imperceptible imaging patterns discovered via multiparametric pattern analysis methods were found to estimate the extent of infiltration and location of future tumor recurrence, paving the way for improved targeted treatment.

Advancing The Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features

Gliomas belong to a group of central nervous system tumors, and consist of various sub-regions. Gold standard labeling of these sub-regions in radiographic imaging is essential for both clinical and computational studies, including radiomic and radiogenomic analyses. Towards this end, we release segmentation labels and radiomic features for all pre-operative multimodal magnetic resonance imaging (MRI) (n=243) of the multi-institutional glioma collections of The Cancer Genome Atlas (TCGA), publicly available in The Cancer Imaging Archive (TCIA). Pre-operative scans were identified in both glioblastoma (TCGA-GBM, n=135) and low-grade-glioma (TCGA-LGG, n=108) collections via radiological assessment. The glioma sub-region labels were produced by an automated state-of-the-art method and manually revised by an expert board-certified neuroradiologist. An extensive panel of radiomic features was extracted based on the manually-revised labels. This set of labels and features should enable i) direct utilization of the TCGA/TCIA glioma collections towards repeatable, reproducible and comparative quantitative studies leading to new predictive, prognostic, and diagnostic assessments, as well as ii) performance evaluation of computer-aided segmentation methods, and comparison to our state-of-the-art method.

Automated Tumor Volumetry Using Computer-Aided Image Segmentation

RATIONALE AND OBJECTIVES Accurate segmentation of brain tumors, and quantification of tumor volume, is important for diagnosis, monitoring, and planning therapeutic intervention. Manual segmentation is not widely used because of time constraints. Previous efforts have mainly produced methods that are tailored to a particular type of tumor or acquisition protocol and have mostly failed to produce a method that functions on different tumor types and is robust to changes in scanning parameters, resolution, and image quality, thereby limiting their clinical value. Herein, we present a semiautomatic method for tumor segmentation that is fast, accurate, and robust to a wide variation in image quality and resolution. MATERIALS AND METHODS A semiautomatic segmentation method based on the geodesic distance transform was developed and validated by using it to segment 54 brain tumors. Glioblastomas, meningiomas, and brain metastases were segmented. Qualitative validation was based on physician ratings provided by three clinical experts. Quantitative validation was based on comparing semiautomatic and manual segmentations. RESULTS Tumor segmentations obtained using manual and automatic methods were compared quantitatively using the Dice measure of overlap. Subjective evaluation was performed by having human experts rate the computerized segmentations on a 0-5 rating scale where 5 indicated perfect segmentation. CONCLUSIONS The proposed method addresses a significant, unmet need in the field of neuro-oncology. Specifically, this method enables clinicians to obtain accurate and reproducible tumor volumes without the need for manual segmentation.