Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Image-localized biopsy mapping of brain tumor heterogeneity: A single-center study protocol

  • Javier C Urcuyo ,

    Contributed equally to this work with: Javier C Urcuyo, Lee Curtin

    Roles Data curation, Formal analysis, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Lee Curtin ,

    Contributed equally to this work with: Javier C Urcuyo, Lee Curtin

    Roles Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Jazlynn M. Langworthy,

    Roles Data curation, Formal analysis, Investigation, Validation, Visualization, Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Gustavo De Leon,

    Roles Conceptualization, Data curation, Investigation, Writing – original draft

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Barrett Anderies,

    Roles Conceptualization, Investigation, Methodology

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Kyle W. Singleton,

    Roles Data curation, Validation, Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Andrea Hawkins-Daarud,

    Roles Data curation, Project administration, Supervision, Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Pamela R. Jackson,

    Roles Project administration, Supervision, Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Kamila M. Bond,

    Roles Data curation, Visualization, Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Sara Ranjbar,

    Roles Writing – review & editing

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Yvette Lassiter-Morris,

    Roles Data curation

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Kamala R. Clark-Swanson,

    Roles Data curation

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Lisa E. Paulson,

    Roles Data curation

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Chris Sereduk,

    Roles Data curation

    Affiliation Department of Cancer Biology, Mayo Clinic, Phoenix, Arizona, United States of America

  • Maciej M. Mrugala,

    Roles Project administration

    Affiliations Department of Neurology, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Oncology, Mayo Clinic, Phoenix, Arizona, United States of America

  • Alyx B. Porter,

    Roles Project administration

    Affiliations Department of Neurology, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Oncology, Mayo Clinic, Phoenix, Arizona, United States of America

  • Leslie Baxter,

    Roles Conceptualization

    Affiliation Department of Neurophysiology, Mayo Clinic, Phoenix, Arizona, United States of America

  • Marcela Salomao,

    Roles Project administration

    Affiliation Department of Pathology, Mayo Clinic, Phoenix, Arizona, United States of America

  • Kliment Donev,

    Roles Project administration

    Affiliation Department of Pathology, Mayo Clinic, Phoenix, Arizona, United States of America

  • Miles Hudson,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Jenna Meyer,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Qazi Zeeshan,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Mithun Sattur,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Devi P. Patra,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Breck A. Jones,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Rudy J. Rahme,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Matthew T. Neal,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Naresh Patel,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Pelagia Kouloumberis,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Ali H. Turkmani,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Mark Lyons,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Chandan Krishna,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Richard S. Zimmerman,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Bernard R. Bendok,

    Roles Investigation

    Affiliation Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America

  • Nhan L. Tran,

    Roles Data curation, Funding acquisition

    Affiliation Department of Cancer Biology, Mayo Clinic, Phoenix, Arizona, United States of America

  • Leland S. Hu,

    Roles Data curation, Funding acquisition, Supervision, Writing – review & editing

    Affiliation Department of Radiology, Mayo Clinic, Phoenix, Arizona, United States of America

  •  [ ... ],
  • Kristin R. Swanson

    Roles Conceptualization, Funding acquisition, Investigation, Project administration, Resources, Supervision, Visualization, Writing – review & editing

    swanson.kristin@mayo.edu

    Affiliations Mathematical NeuroOncology Lab, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Neurosurgery, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Cancer Biology, Mayo Clinic, Phoenix, Arizona, United States of America, Department of Radiation Oncology, Mayo Clinic, Phoenix, Arizona, United States of America

  • [ view all ]
  • [ view less ]

Abstract

Brain cancers pose a novel set of difficulties due to the limited accessibility of human brain tumor tissue. For this reason, clinical decision-making relies heavily on MR imaging interpretation, yet the mapping between MRI features and underlying biology remains ambiguous. Standard (clinical) tissue sampling fails to capture the full heterogeneity of the disease. Biopsies are required to obtain a pathological diagnosis and are predominantly taken from the tumor core, which often has different traits to the surrounding invasive tumor that typically leads to recurrent disease. One approach to solving this issue is to characterize the spatial heterogeneity of molecular, genetic, and cellular features of glioma through the intraoperative collection of multiple image-localized biopsy samples paired with multi-parametric MRIs. We have adopted this approach and are currently actively enrolling patients for our ‘Image-Based Mapping of Brain Tumors’ study. Patients are eligible for this research study (IRB #16–002424) if they are 18 years or older and undergoing surgical intervention for a brain lesion. Once identified, candidate patients receive dynamic susceptibility contrast (DSC) perfusion MRI and diffusion tensor imaging (DTI), in addition to standard sequences (T1, T1Gd, T2, T2-FLAIR) at their presurgical scan. During surgery, sample anatomical locations are tracked using neuronavigation. The collected specimens from this research study are used to capture the intra-tumoral heterogeneity across brain tumors including quantification of genetic aberrations through whole-exome and RNA sequencing as well as other tissue analysis techniques. To date, these data (made available through a public portal) have been used to generate, test, and validate predictive regional maps of the spatial distribution of tumor cell density and/or treatment-related key genetic marker status to identify biopsy and/or treatment targets based on insight from the entire tumor makeup. This type of methodology, when delivered within clinically feasible time frames, has the potential to further inform medical decision-making by improving surgical intervention, radiation, and targeted drug therapy for patients with glioma.

Introduction

Brain tumors are uniquely challenging due to the limited access intrinsic to their location delaying progress in understanding and developing effective treatments. For example, the most aggressive primary brain cancer, glioblastoma (GBM), is particularly well known for its short median survival from diagnosis, and once systemic cancers reach the stage of brain metastases, patients typically experience very poor outcomes [13]. Surgery remains one of the most notable treatments for all brain tumors. In some cases, the benefits go past debulking, as surgery also provides an opportunity to collect tissue and elucidate genetic, transcriptomic and environmental factors that may act as treatment targets or prognostic markers. However, capturing the full heterogeneity within brain tumors remains impossible given the eloquent brain tissue in which these tumors live. The protocol introduced is intended to lead to a large dataset of image localized biopsies that will be able to answer some of these key questions surrounding the heterogeneity of brain tumors.

The landscape of each brain tumor is ever changing, and there are limited opportunities for tissue collection, so it is vital that these limited surgical sampling opportunities are capitalized on. Diffusely invasive gliomas are particularly challenging in this regard as there are always tumor cells left behind in the brain following any surgical intervention. GBM, the most aggressive form of glioma, is the most common primary brain malignancy among adults [4]. Despite aggressive medical intervention consisting of maximal surgical resection followed by concurrent chemoradiation and adjuvant temozolomide, this disease remains uniformly fatal, with a median survival rate of 14–16 months in patients with newly-diagnosed GBM [5, 6] and 5–7 months in patients with recurrent GBM [4]. This poor prognosis is often a result of a major hallmark of GBM: profound intratumoral heterogeneity that contributes to treatment resistance and tumor recurrence. The presence of metastatic brain lesions also corresponds to a poor prognosis. These lesions have common alterations compared to their primary counterparts [7]. However, there is also some evidence pointing towards diversity within and between metastatic brain lesions [8, 9]. We collect metastatic tumor tissue to understand this heterogeneity and determine how the different metastatic tissue features may present on imaging.

Given the nexus of known intratumoral heterogeneity and the limitations of access to brain tumor tissue, there is an urgent need to leverage imaging to better inform our understanding of the biology at play across patients and within each patient’s tumor. For example, the molecular composition of tumors is important, as different tumor cell subpopulations within and between patients can have different treatment sensitivities and implications for survival. For instance, IDH1/IDH2 and MGMT are prognostic markers of survival for gliomas [5, 6]. Notably, mutation in IDH is thought to be uniform throughout a patient tumor [10, 11], while MGMT status can vary with treatment status and location [1214]. Other genetic alterations can arise, with multiple phenotypes present within the same tumor. For instance, the epidermal growth factor receptor (EGFR) gene can differ significantly in different regions of the tumor [1518]. Similarly, EGFR-targeted therapies have mixed responses in patients [19, 20]. Bulk transcriptional studies of GBM, derived from The Cancer Genome Atlas data, have revealed tumor subtypes that are associated with specific genetic alterations and patient survival [21, 22]. These subtypes have been recapitulated on the level of individual cells, and profound genomic and phenotypic variability has been identified within individual patients [2225]. These heterogeneities in key markers suggest that, although sufficient for diagnosis, clinical samples likely do not represent the full genetic and transcriptomic complexity of each tumor. It is important to have an accurate representation of the tumor throughout the clinical course of care, but opportunities for tissue collection are limited. This lack of tissue access has led to a heavy reliance on imaging to assess important clinical benchmarks such as surgical outcome, tumor size, and treatment response.

Due to its ubiquitous use in the clinic, there has been a long history of using MRI to predict the extent and biological heterogeneity of brain tumors [2640]. Although they have shown promise, validating these predictions has been extremely difficult without access to local tissue samples. Image-localized biopsies, such as those collected here, have the potential to be used in developing and validating such spatial mathematical models as well as emboldening newer approaches such as radiomics, where machine-learning models are trained to connect high throughput quantitative image features to the composition of brain tumor tissue [16, 18, 4146]. Given limited access to brain tumor tissue, typical radiomics studies aim to connect image features from tumors to global characteristics of the tumors such as malignancy, IDH1 status and grade [47] whereas image-localized biopsies provide an opportunity to embrace intratumoral heterogeneity. This same quantitative approach (image analysis meets tissue analysis meets machine learning) can and has been used to link local imaging patterns at biopsy sites with the composition of GBM tissue of said biopsies [16, 18, 41, 45, 46]. Given that many clinical decisions in the treatment and monitoring of GBMs are based on MRI, connecting GBM tissue composition to MRI offers the potential to capture the evolving tumor landscape and cellular populations noninvasively, leading to more informed clinical decisions to benefit the patient. As we collect more image-localized biopsies, this dataset becomes more representative of the vast range of possible tumor compositions and imaging appearances, giving machine learning models the opportunity to become more robust. Such robust models have the potential to predict important tissue features of GBM through imaging alone on a patient-specific spatial basis, which would arm clinicians with the knowledge to provide more nuanced treatment and better stratify patients for clinical trials.

This clinical protocol’s sample collection began in October 2017, and is ongoing. Collection of samples will continue in efforts to grow statistical power as long as we are able to conduct this research. This protocol has no impact on consented patients’ clinical care, nor alters anything within the clinical milieu.

Imaging-defined tumor regions drive clinical decisions

Currently, maximal safe surgical resection is the clinical gold standard. However, this is often determined with postoperative magnetic resonance imaging (MRI), particularly the T1-weighted MRI with gadolinium contrast (T1+C) [48, 49]. Tissue sampling creates significant challenges for studying the clonal diversity of tumors such as GBM. T1+C MRI represents the clinical standard for neuronavigation and routinely guides surgical biopsies and resection from the MRI enhancing core. In fact, a clinically defined “gross total resection”, for high grade glioma, is defined as removal of the complete T1+C abnormality. Unfortunately, biopsies from this contrast-enhancing (CE) tumor region alone fails to address the diverse molecularly-distinct subpopulations that extend into the surrounding non-enhancing (NE) parenchyma, which is visible on T2-weighted/Fluid-Attenuated Inversion Recovery (T2W/FLAIR) MRI [16]. These generally unresected NE tumor regions contribute to tumor recurrence and can have different cellular compositions and genetic signatures to that of enhancing regions [25, 50, 51]. Furthermore, T1+C MRI fails to localize cancer in the surrounding NE tumor region during radiation treatment (RT) planning, as non-tumoral edema typically appears visually indistinguishable from NE tumor. Most radiation oncologists must apply submaximal doses across the entire T2W/FLAIR volume, which delivers unnecessary radiation to the normal brain and risks undertreating NE tumor.

Advanced multiparametric imaging provides deeper insights into tumor biology

Imaging techniques, such as advanced MRI, can quantitatively characterize tumor-induced physiological processes in the NE region of GBM. Unlike surgical sampling, MRI captures the entire brain organ and thus could provide insight into the tumor extent even in unresected NE regions and their associated biophysical features that may be detectable on MRI (Fig 1). On T1+C MRI, enhancement indicates regions of disrupted blood brain barrier (BBB), while signal demarcates regions of high water content and tumoral edema in T2W/FLAIR. Other advanced MRI features may reflect tumor cell density on diffusion-weighted imaging (DWI) [52], white matter infiltration on diffusion tensor imaging (DTI) [53, 54], and microvessel morphology on Dynamic Susceptibility-weighted contrast-enhanced perfusion MRI (DSC-pMRI) [55]. In addition, signal intensity values on structural MRIs are spatial representations of soft tissue anatomy. The textural patterns of neighboring voxel intensities provide further insight towards the potential tissue microstructure and phenotypic heterogeneity within the local microenvironment [56, 57]. These complementary MRI features offer potential biomarkers of underlying genomic and transcriptomic status, and have been previously correlated with molecular profiles of GBM [16, 5865]. Further, quantification of the interactions amongst molecularly-distinct subpopulations, cellular subpopulation compositions and/or their diversity in the NE tumor region (often left behind following surgical interventions) can help improve future treatment strategies (such as adaptive therapy), under the realm of individualized oncology [6668].

thumbnail
Fig 1. Multiparametric MRI techniques and contrasts.

Listed are the 8 different MRI sequences used in this study, along with their corresponding physiological representations. T1Gd = T1W signal increase on post-contrast imaging; FLAIR = Fluid-Attenuated Inversion Recovery; rCBV = Relative cerebral blood volume; DSC-pMRI = Dynamic susceptibility contrast perfusion MRI; EPI+C = T2*W signal loss (i.e., negative enhancement at ~5min post injection of 0.1 mmol/kg preload contrast injection); DWI = Diffusion weighted imaging; DTI = Diffusion tensor imaging; MD = Mean diffusivity (equivalent to apparent diffusion coefficient); FA = Fractional anisotropy.

https://doi.org/10.1371/journal.pone.0287767.g001

Materials and methods

The ‘Image-Based Mapping of Brain Tumors’ clinical protocol (IRB# 16–002424) is a minimal-risk study that aims to further characterize intratumoral heterogeneity of brain tumors. Here, we outline our study collecting multiple image-guided biopsy samples as part of standard surgical approaches. We aim to collect samples dispersed in and around the abnormality seen on T1Gd MRI to allow for cross-annotation for relevant MRI features. As these image-localized biopsies are collected during routine surgical interventions, this study does not alter the course of treatment for consented patients, but we hope that findings using this data will ultimately help patients in the future. We present an example set of biopsy locations for a single surgery in Fig 2. The data gathered through this study will be useful for countless research directions. The study’s primary objectives are to identify relationships between imaging and a number of important tumor features, such as tumor cell density, genetic status, transcriptomic status, and molecular status. Secondary objectives include investigating associations amongst imaging, radiation dosimetry, tumor recurrence/treatment effect, and clinical outcome. We foresee the use of statistical methods such as repeated measures correlations [69] and mixed effects models [70] to test for gene correlations and differences between imaging regions, Cox mixed effects models [71] to determine impact on clinical outcome, and supervised machine learning models such as regressors or classifiers to connect tumor composition (cellular, genetic, transcriptomics) to the more readily available imaging data.

thumbnail
Fig 2. Illustrative example of biopsy locations annotated relative to imaging abnormalities.

Here we show an example case where 9 image-localized biopsies were acquired from a patient. (A) An example screenshot from the intraoperative neuronavigation system of a non-enhancing (NE) biopsy sample. (B) An example screenshot from the intraoperative neuronavigation system of a contrast-enhancing (CE) biopsy sample. (C) A screenshot from intraoperative navigation is shown, highlighting the translation to a full three-dimensional rendering of segmented imaging abnormalities and biopsy locations. The segmented CE (dark blue) and NE (light blue) abnormalities are shown with biopsy locations (red). (D-F) MRI planes overlaid with segmentations and biopsy locations, highlighting the representative spread of biopsy locations that we can achieve. Although only the CE and NE regions are outlined here, we collect other multiparametric MRIs as well for the patients as shown in Fig 1.

https://doi.org/10.1371/journal.pone.0287767.g002

Inclusion and exclusion criteria

Inclusion criteria for patient participation in the study are as follows:

  • Age between 18 and 99 (inclusive)
  • Undergoing diagnostic biopsy and/or surgical resection for a brain lesion

Exclusion criteria for patient participation in the study are as follows:

  • Insufficient renal function: eGFR < 60 mg/min/1.72m2 [72]
  • Allergy to Gadolinium (Gd)
  • Pregnant or nursing
  • History of hemolytic anemia or asthma
  • Inability to obtain informed written consent

Workflow

Before surgery.

Patients are identified through the institutional electronic medical records system (i.e., Epic Systems). Clinical schedules for MR scanners and operating rooms are checked daily for eligible participants. A study-specific multi-parametric imaging protocol is acquired prior to surgery for eligible patients. The imaging protocol includes T1-weighted (T1W), T1-weighted with gadolinium contrast (T1Gd), T2-weighted (T2W), T2W-fluid-attenuated inversion recovery (T2-FLAIR), diffusion tensor imaging (DTI), and dynamic susceptibility contrast (DSC) perfusion MRI (Fig 1). These imaging sequences are acquired prior to the collection of image-guided biopsy samples and, in some instances, following the patient’s diagnosis or treatment. In addition to MRI, other imaging modalities may be collected and reviewed, these include computed tomography (CT) and positron emission tomography (PET) imaging data collected as part of clinical practice or in conjunction with other imaging-based protocols.

Blood is collected in up to two 10mL ethylenediamine tetraacetic acid (EDTA) tubes and the buffy coat is frozen. This blood collection is used for germline DNA comparisons to the respective tumors and occurs during a standard of care visit. If the patient has multiple standard-of-care lab visits, we may collect at any of those unless the subject has withdrawn informed consent.

Urine is collected and stored at 4ºC to assess biomarkers (free circulating DNA/RNA) associated with therapy response. This may also be collected at any standard of care lab visits, as long as the subject has not withdrawn informed consent. Table 1 shows the timeline for the collection of all sections of this protocol.

thumbnail
Table 1. General schedule of events for this research study from patient identification to routine follow-up.

Patient information beyond routine follow-up can also be collected. No additional modifications are made to the standard clinical care trajectory unless the patient has another surgery.

https://doi.org/10.1371/journal.pone.0287767.t001

During surgery.

During routine surgery, biopsies are obtained, and the MRI location is recorded using an intraoperative neuronavigation system (Medtronic StealthStation 8, Minneapolis, MN) with additional screenshots taken (such as that in Fig 3) to further validate the coordinate location. At the beginning of this study, biopsies were frozen by the Surgical Pathology department. However, to reduce time to freezing of the tissue, flash-freezing was implemented in the operating room (OR) with Surgical Pathology as a backup, if necessary. This has significantly improved our success in achieving freezing tissue within 5 minutes of surgical extraction (S1 Fig).

thumbnail
Fig 3. A schematic of our per-patient biopsy collection workflow.

Additionally, patients may have samples collected at subsequent surgeries [7377].

https://doi.org/10.1371/journal.pone.0287767.g003

We collect multiple spatially-annotated stereotactic biopsies from the across the diversity of MRI-definable tumor regions in patients with glioma and metastatic brain tumors with a goal of 8 biopsies per surgery with a target of ≥250 mg/biopsy (see Fig 4). Tumor grade is assessed by a pathologist using corresponding clinical biopsy samples in line with standard conventions [78, 79].

thumbnail
Fig 4. Overview of image-localized biopsy histology distribution.

(Top) Total number of surgeries with image-localized biopsies collected, broken down by grade (total of 183 surgeries). (Middle) Number of samples per surgery, broken down by grade. (Bottom) Total samples collected per grade (total of 1140 image-localized samples). Met = Metastatic brain tumor.

https://doi.org/10.1371/journal.pone.0287767.g004

Originally, biopsies were only flash-frozen in order to conduct whole-exome and RNA sequencing on the samples. Since December of 2022, we also began submerging tissue in formalin. At first, this was done by the surgeon collecting extra samples, 2 at each coordinate location. However, this overstretched the clinical workflow. In consultation with clinical staff, it was decided that one sample will be collected from each coordinate location, and if the sample is large enough the research team will split them in two; one is then flash frozen, and the other formalin fixed. This way, we have more information about the biology in each of these areas, without any increased risk to the patient or workload on the clinical staff. We have so far collected a total of 45 formalin-fixed samples from 8 surgeries.

Screenshots taken for each research sample are extracted from the neuronavigation system and imported into our IRB-compliant database PatientView (Fig 3). Samples are time-stamped and designated an alphabetical letter to facilitate matching each sample to its corresponding screenshot and clinical notes in the electronic medical record. Samples collected are also cross-referenced with the Surgical Pathology department to confirm the number of biopsies collected, label accuracy, and overall quality assurance.

After surgery.

Imaging data is collected, coded, processed, and matched with the genetic and molecular data obtained from each biopsy. Post-processing analyses include registration, normalization, inhomogeneity correction, and feature extraction as described elsewhere [16, 41, 46]. Planned studies include the development of statistical models between imaging, various tissue characterizations (e.g., copy number variants, transcriptomic signatures, immunohistochemistry etc.), and clinical outcomes.

Each sample is delivered to the Surgical Pathology department after being flash-frozen in liquid nitrogen. The samples are stored in a -80ºC freezer until subsequent processing. Flash-frozen tissue is retrieved and embedded frozen in optimal cutting temperature (OCT) compound. Tissue is sectioned (e.g., target 10μm with the goal of up to 20 slides) in -20ºC cryostat (e.g., Microm-HM-550) utilizing a microtome blade. In the event of excess archived formalin-fixed paraffin-embedded (FFPE) tissue collected per the standard clinical protocol; retrospective tissue may be obtained (under the discretion of a neuropathologist) to undergo further tissue analysis (we aim to collect up to 100μm in FFPE scrolls). All specimens are also stained with hematoxylin and eosin (H&E) and reviewed by a neuropathologist to quantify tumor content. Tissue specimens may be submitted for subsequent genetic, molecular, or epigenetic analysis, including, but not exclusive to, next-generation sequencing, array-based comparative genomic hybridization (aCGH), exome sequencing, methylation analysis, and RNA sequencing.

Data abstraction, management, and availability.

Details of the patient’s clinical course, treatments, MRI images, pathologies, and treatment response are abstracted from the medical records system by IRB-approved staff and an anonymized data repository is distributed to the rest of the team for data analysis. Data is made accessible to all IRB-approved staff, but only the principal investigators, study coordinators, and research assistants have full rights to update data in our main data repository as they are responsible for ensuring the data quality and accuracy.

Biopsy-related data, including time of collection, specimen ID, neuronavigation information, and other notes are collected by two to three researchers during surgery in physical journals. These journals contain coded patient IDs with no patient health information. This data is transferred to a password-protected document with all relevant data for each patient, which is stored on a secure server that can only be accessed by IRB-approved researchers. Biopsy location data and image information is taken directly from the neuronavigation system using plans and time-stamped screenshots as previously described. The image data and screenshots are kept in deidentified patient folders on the secure server. Patient identifiers, demographic information, and sample data are collated in the main data repository.

Research data that documents, supports, and validates research findings will be made available after the main findings from the final research data set have been accepted for publication. We will provide these in the form of csv files. In addition, if requested, tissue and imaging data may be made available for sharing to qualified parties by the technology transfer office as soon as is reasonably possible, so long as such a request does not compromise intellectual property interests, interfere with publication, invade subject privacy, or betray confidentiality. Data that are shared will include standards and notations needed to interpret the data, following commonly accepted practices. All data from tissue processed from this cohort will be made available through the MOSAIC consortium we have formed around this important novel intratumoral heterogeneity initiative: www.BrainTumorMOSAIC.org. Additional data requests may be initiated by contacting the Swanson lab through our website: www.MathematicalNeuroOncology.org.

Ethical considerations and declarations

This study was approved by the Mayo Clinic IRB on 1/9/2017 with approval number 16–002424. All participants presented in this work have undergone written informed consent. The principal investigators are responsible for ensuring the IRB-approved study protocol is followed and for reporting any adverse events. The study protocol is reviewed for renewal annually.

Safety considerations.

Our study is not therapeutically interventional. This is a minimal-risk study as procedures are in line with standard clinical activities with only the possible addition of advanced imaging. The safety of the patient is our top priority and is monitored by the neurosurgeon as surgery is performed.

Patient demographics and sample counts

As of April 15th 2023, this study has enrolled 203 patients with 184 surgeries that resulted in the successful collection of biopsy samples (183 of which collected image-localized biopsies). A total of 1175 biopsies have been collected of which 1140 are image localized, with a mean of 6.38 samples per surgery. The mean age of these patients was 57.5 years old (ranging from 18–91). The self-reported sex breakdown is 113 males and 90 females with mean ages of 58.2 years (20–84) and 56.6 years (18–91), respectively. The self-reported racial and ethnic distribution of the cohort is largely white and non-Hispanic/Latino individuals (Fig 5, analogous sex-specific figures in S2 Fig). The samples collected in this study were largely acquired from patients with glioma and brain metastasis, although 10 patients had another type of brain lesion. We collected image-localized biopsies from 134 glioma surgeries, of which: 4 were grade 1, 16 were grade 2, 17 were grade 3 and 97 were grade 4. We collected image-localized biopsies from 39 surgeries for patients with cancers that had metastasized to the brain. See Fig 6 for further breakdowns such as tumor status and biopsy counts (analogous sex-specific figures can be found in S3 and S4 Figs).

thumbnail
Fig 5. Self-reported race (top) and ethnicity (bottom) in the patient cohort, which consists predominantly of patients reporting as white and not Hispanic/Latino.

See S2 Fig for sex-specific breakdowns.

https://doi.org/10.1371/journal.pone.0287767.g005

thumbnail
Fig 6. Flowchart for enrollment, starting from the number of patients (N) who were approached for the study, to surgeries (S) and total number of biopsies collected (bx), stratified by grade and tumor status.

See S3 and S4 Figs for sex-specific breakdowns. *One case included in the recurrent grade III group underwent a grade transformation.

https://doi.org/10.1371/journal.pone.0287767.g006

Discussion

As with any clinical study, there are many potential limitations that must be considered. This study aims to retrieve multiple biopsies from a variety of locations spanning the diversity of MRI-defined tumor regions. For example, in glioma it is particularly important that we biopsy from the diffusely invaded normal appearing brain parenchyma, as these samples will be more similar to the unresected invasive margins that repopulate the disease. While we have a goal of 8 samples per surgery, in some cases this sampling scheme is not possible. In other cases, we may not be able to collect from all imaging regions. The operating neurosurgeon is responsible for evaluating the safety of biopsy collection and assessing any potential risk that may result from such collection. As a result, there is notable variability in the number and size of the collected specimens; however, this has helped define our study as a reasonable and pragmatic collection protocol.

Image-localization poses additional challenges. First, the neuronavigation system requires preoperative patient registration, resulting in a measurable registration error during surgery. Second, this process is reliant on preoperative imaging, which must be ordered by clinicians and, in some scenarios, is obtained up to three weeks prior to surgery. Further, such static images cannot account for intraoperative brain shift, a phenomenon where the brain fills the space of resected tissue. Unfortunately, this issue remains a universal challenge for image-localization during neurological surgery, and there is currently no standard procedure in place to accurately measure this change intraoperatively. These are all important considerations for image-based modeling. In addition, different MRI machines with different techs and field strengths can lead to varying imaging features, even within the same patient. Rather than alter clinical workflow, we must find ways (such as normalization techniques) to bring these images together into a comparable space for radiomics models. We have provided a framework for such radiomic model generation in recent work [80].

Another limitation is in the uniformity of available imaging and the related difficulty of consenting patients promptly. Patients are identified through our institution’s patient scheduling systems (currently in EPIC, the electronic medical records system utilized at Mayo Clinic). Specifically, MRI and OR schedules are monitored for potential patients for the study. However, emergent patients who undergo imaging at short notice and immediately proceed with surgery may be missed or, if consented, may not have received the entire protocol of imaging before surgery. Since biopsy collections require research personnel, surgeries may also be missed if researchers’ schedules do not permit attendance. Further, surgeries are not attended if there is increased risk to the patient (e.g., patient safety concerns from the neurosurgeon) or the research team (i.e., active COVID-19 diagnosis).

Although this data collection comes with challenges, many of these are typical during the integration of research into a clinical workflow. Neuronavigation has proven to be a useful clinical tool for surgical planning and intraoperative guidance. By utilizing neuronavigation for research, we can attain much greater insight into the inter- and intra-patient spatial heterogeneity of brain tumors and its imaging presentation. These insights may provide therapeutic targets for clinical trials, and we hope one day they will improve treatment options.

For all research approaches, we want our dataset to be both representative of each tumor and the patient cohort we ultimately aim to serve. Of course, more data is always desirable, particularly for radiomics models, but it is unrealistic to expect the clinical team to collect all resected tissue as image-localized biopsies. With this protocol, we’ve mindfully struck a balance between representative data, minimal patient risk, and minimal impact on the clinical workflow. It is our hope that spatial localization of brain tumor biopsies becomes standard in the clinic to increase our understanding of these tumors, inform clinical care, and ultimately improve prognosis for patients.

Supporting information

S1 Fig. Time taken to freeze biopsy samples.

For 1109/1130 flash-frozen samples, we have tracked the time from extraction to freezing. Switching from flash-freezing in pathology (N = 597) to inside the operating room (OR, N = 512) has allowed us to significantly reduce time between surgical extraction and freezing (linear mixed effects model with surgery ID as a nested random variable, p<0.0001). In doing so, we reach the goal of freezing biopsies in a 5-minute window (red dashed line) more consistently.

https://doi.org/10.1371/journal.pone.0287767.s001

(PDF)

S2 Fig. Self-reported race and ethnicity broken down by self-reported sex of patients.

Each sex consists predominantly of patients who identified as white, and patients who identified as not Hispanic/Latino.

https://doi.org/10.1371/journal.pone.0287767.s002

(PDF)

S3 Fig. Counts breakdown for female patients.

The breakdown of patients approached and consented, as well as tumor type, grade and treatment status for patients who self-identified as female. *One case included in the recurrent grade III group underwent a grade transformation.

https://doi.org/10.1371/journal.pone.0287767.s003

(PDF)

S4 Fig. Counts breakdown for male patients.

The breakdown of patients approached and consented, as well as tumor type, grade and treatment status for patients who self-identified as male.

https://doi.org/10.1371/journal.pone.0287767.s004

(PDF)

Acknowledgments

One might say “it takes a village” to achieve a successful protocol of this scope. This manuscript focuses on the establishment of our protocol, current/recent workflows, processes, and staffing that we have implemented. As such, the authors are truly grateful to the many additional contributions by other prior and current members of the biopsy collection team, including but not limited to:

Sample Collection: Ariana Afshari, Spencer Bayless, Rhea Carlson, Ryan Hess, Julia Lorence, Reyna Patel, Lucas Paulson, Sebastian Velez, Ryan Eghlimi, Angad Beniwal, Gabe LaFond

Clinical Research Coordinators: Jessica Bauer, Regina Becker, Lysette Elsner, Crystal Harris, Morgan Hatlestead, Ashley Napier

Surgical Technologists: Darby Black, Heather Boles, Heather Hull, Angela Melloni, Jose Rocha, Briana Rodriguez

Clinical Nursing Staff: Nora J. Shaefer, Kaylee Curley, Hans Leitner DeCarlo, Deanna Dusek, Regina Formentin, Braden Hall, Nathan Nitzky, Melanie Parativo, Kenneth Rooth, Krystle Short, Matt Zumwalt Hulverson

Stealth Techs: Cara Burkholder, Albana Spahiu

Interventional Radiology Technologists: Jason Billings, Joel Dela Cruz, Emily Lange, Brandy Streeter, Patrick Vance

Surgical Pathology Staff: Bradley Atkins, Debra Keith, Phillip Hogan Jr, Rachel Matey, Lori Miller, Ashley Polzin, Jessica Williams

Clinical data abstraction: Sandra K. Johnston

Clinical research database integration/administration: Scott Whitmire

Other project management and design: Susan C. Massey

Image analysis: Cassandra R. Rickertsen and past and current members of the Image Analysis Team.

References

  1. 1. Hall WA, Djalilian HR, Nussbaum ES, Cho KH. Long-term survival with metastatic cancer to the brain. Med Oncol. 2000;17: 279–286. pmid:11114706
  2. 2. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJB, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005;352: 987–996. pmid:15758009
  3. 3. Stupp R, Taillibert S, Kanner A, Read W, Steinberg D, Lhermitte B, et al. Effect of Tumor-Treating Fields Plus Maintenance Temozolomide vs Maintenance Temozolomide Alone on Survival in Patients With Glioblastoma: A Randomized Clinical Trial. JAMA. 2017;318: 2306–2316. pmid:29260225
  4. 4. Ostrom QT, Cioffi G, Gittleman H, Patil N, Waite K, Kruchko C, et al. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2012–2016. Neuro Oncol. 2019;21: v1–v100. pmid:31675094
  5. 5. Molenaar RJ, Verbaan D, Lamba S, Zanon C, Jeuken JWM, Boots-Sprenger SHE, et al. The combination of IDH1 mutations and MGMT methylation status predicts survival in glioblastoma better than either IDH1 or MGMT alone. Neuro Oncol. 2014;16: 1263–1273. pmid:24510240
  6. 6. Kalkan R, Atli Eİ, Özdemir M, Çiftçi E, Aydin HE, Artan S, et al. IDH1 mutations is prognostic marker for primary glioblastoma multiforme but MGMT hypermethylation is not prognostic for primary glioblastoma multiforme. Gene. 2015;554: 81–86. pmid:25455102
  7. 7. Brastianos PK, Carter SL, Santagata S, Cahill DP, Taylor-Weiner A, Jones RT, et al. Genomic Characterization of Brain Metastases Reveals Branched Evolution and Potential Therapeutic Targets. Cancer Discov. 2015;5: 1164–1177. pmid:26410082
  8. 8. Ali S, Górska Z, Duchnowska R, Jassem J. Molecular Profiles of Brain Metastases: A Focus on Heterogeneity. Cancers. 2021;13: 2645. pmid:34071176
  9. 9. Liu Y, Ye G, Huang L, Zhang C, Sheng Y, Wu B, et al. Single-cell transcriptome analysis demonstrates inter-patient and intra-tumor heterogeneity in primary and metastatic lung adenocarcinoma. Aging. 2020;12: 21559–21581. pmid:33170151
  10. 10. Juratli TA, Kirsch M, Robel K, Soucek S, Geiger K, von Kummer R, et al. IDH mutations as an early and consistent marker in low-grade astrocytomas WHO grade II and their consecutive secondary high-grade gliomas. J Neurooncol. 2012;108: 403–410. pmid:22410704
  11. 11. Duncan CG, Barwick BG, Jin G, Rago C, Kapoor-Vazirani P, Powell DR, et al. A heterozygous IDH1R132H/WT mutation induces genome-wide alterations in DNA methylation. Genome Res. 2012;22: 2339–2355. pmid:22899282
  12. 12. Parkinson JF, Wheeler HR, Clarkson A, McKenzie CA, Biggs MT, Little NS, et al. Variation of O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation in serial samples in glioblastoma. J Neurooncol. 2007;87: 71–78. pmid:18004504
  13. 13. Dunn J, Baborie A, Alam F, Joyce K, Moxham M, Sibson R, et al. Extent of MGMT promoter methylation correlates with outcome in glioblastomas given temozolomide and radiotherapy. Br J Cancer. 2009;101: 124–131. pmid:19536096
  14. 14. Brandes AA, Franceschi E, Paccapelo A, Tallini G, De Biase D, Ghimenton C, et al. Role of Methylation Status at Time of Diagnosis and Recurrence for Patients with Glioblastoma: Clinical Implications. Oncologist. 2017;22: 432–437.
  15. 15. Szerlip NJ, Pedraza A, Chakravarty D, Azim M, McGuire J, Fang Y, et al. Intratumoral heterogeneity of receptor tyrosine kinases EGFR and PDGFRA amplification in glioblastoma defines subpopulations with distinct growth factor response. Proc Natl Acad Sci U S A. 2012;109: 3041–3046. pmid:22323597
  16. 16. Hu LS, Ning S, Eschbacher JM, Baxter LC, Gaw N, Ranjbar S, et al. Radiogenomics to characterize regional genetic heterogeneity in glioblastoma. Neuro Oncol. 2017;19: 128–137. pmid:27502248
  17. 17. Hu LS, Hawkins-Daarud A, Wang L, Li J, Swanson KR. Imaging of intratumoral heterogeneity in high-grade glioma. Cancer Lett. 2020. pmid:32112907
  18. 18. Hu LS, Wang L, Hawkins-Daarud A, Eschbacher JM, Singleton KW, Jackson PR, et al. Uncertainty quantification in the radiogenomics modeling of EGFR amplification in glioblastoma. Sci Rep. 2021;11: 3932. pmid:33594116
  19. 19. Hegi ME, Rajakannu P, Weller M. Epidermal growth factor receptor: a re-emerging target in glioblastoma. Curr Opin Neurol. 2012;25: 774–779. pmid:23007009
  20. 20. Eskilsson E, Røsland GV, Solecki G, Wang Q, Harter PN, Graziani G, et al. EGFR heterogeneity and implications for therapeutic intervention in glioblastoma. Neuro Oncol. 2018;20: 743–752. pmid:29040782
  21. 21. Verhaak RGW, Hoadley KA, Purdom E, Wang V, Qi Y, Wilkerson MD, et al. Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell. 2010;17: 98–110. pmid:20129251
  22. 22. Wang Q, Hu B, Hu X, Kim H, Squatrito M, Scarpace L, et al. Tumor Evolution of Glioma-Intrinsic Gene Expression Subtypes Associates with Immunological Changes in the Microenvironment. Cancer Cell. 2017;32: 42–56.e6. pmid:28697342
  23. 23. Patel AP, Tirosh I, Trombetta JJ, Shalek AK, Gillespie SM, Wakimoto H, et al. Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science. 2014;344: 1396–1401. pmid:24925914
  24. 24. Neftel C, Laffy J, Filbin MG, Hara T, Shore ME, Rahme GJ, et al. An Integrative Model of Cellular States, Plasticity, and Genetics for Glioblastoma. Cell. 2019;178: 835–849.e21. pmid:31327527
  25. 25. Gill BJ, Pisapia DJ, Malone HR, Goldstein H, Lei L, Sonabend A, et al. MRI-localized biopsies reveal subtype-specific differences in molecular and cellular composition at the margins of glioblastoma. Proc Natl Acad Sci U S A. 2014;111: 12550–12555. pmid:25114226
  26. 26. Tripathi S, Vivas-Buitrago T, Domingo RA, Biase GD, Brown D, Akinduro OO, et al. IDH-wild-type glioblastoma cell density and infiltration distribution influence on supramarginal resection and its impact on overall survival: a mathematical model. J Neurosurg. 2021; 1–9. pmid:34715662
  27. 27. Morris B, Curtin L, Hawkins-Daarud A, Hubbard ME, Rahman R, Smith SJ, et al. Identifying the spatial and temporal dynamics of molecularly-distinct glioblastoma sub-populations. Math Biosci Eng. 2020;17: 4905–4941. pmid:33120534
  28. 28. Baldock AL, Ahn S, Rockne R, Johnston S, Neal M, Corwin D, et al. Patient-specific metrics of invasiveness reveal significant prognostic benefit of resection in a predictable subset of gliomas. PLoS One. 2014;9: e99057. pmid:25350742
  29. 29. Baldock AL, Yagle K, Born DE, Ahn S, Trister AD, Neal M, et al. Invasion and proliferation kinetics in enhancing gliomas predict IDH1 mutation status. Neuro Oncol. 2014;16: 779–786. pmid:24832620
  30. 30. Swanson KR, Bridge C, Murray JD, Alvord EC Jr. Virtual and real brain tumors: using mathematical modeling to quantify glioma growth and invasion. J Neurol Sci. 2003;216: 1–10. pmid:14607296
  31. 31. Swanson KR, Alvord EC Jr, Murray JD. A quantitative model for differential motility of gliomas in grey and white matter. Cell Prolif. 2000;33: 317–329. pmid:11063134
  32. 32. Swanson KR, Rockne RC, Claridge J, Chaplain MA, Alvord EC Jr, Anderson ARA. Quantifying the role of angiogenesis in malignant progression of gliomas: in silico modeling integrates imaging and histology. Cancer Res. 2011;71: 7366–7375. pmid:21900399
  33. 33. Neal ML, Trister AD, Cloke T, Sodt R, Ahn S, Baldock AL, et al. Discriminating survival outcomes in patients with glioblastoma using a simulation-based, patient-specific response metric. PLoS One. 2013;8: e51951. pmid:23372647
  34. 34. Neal ML, Trister AD, Ahn S, Baldock A, Bridge CA, Guyman L, et al. Response classification based on a minimal model of glioblastoma growth is prognostic for clinical outcomes and distinguishes progression from pseudoprogression. Cancer Res. 2013;73: 2976–2986. pmid:23400596
  35. 35. Swan A, Hillen T, Bowman JC, Murtha AD. A Patient-Specific Anisotropic Diffusion Model for Brain Tumour Spread. Bull Math Biol. 2018;80: 1259–1291. pmid:28493055
  36. 36. Elazab A, Wang C, Gardezi SJS, Bai H, Hu Q, Wang T, et al. GP-GAN: Brain tumor growth prediction using stacked 3D generative adversarial networks from longitudinal MR Images. Neural Netw. 2020;132: 321–332. pmid:32977277
  37. 37. Subramanian S, Gholami A, Biros G. Simulation of glioblastoma growth using a 3D multispecies tumor model with mass effect. J Math Biol. 2019;79: 941–967. pmid:31127329
  38. 38. Curtin L, Hawkins-Daarud A, Porter AB, van der Zee KG, Owen MR, Swanson KR. A Mechanistic Investigation into Ischemia-Driven Distal Recurrence of Glioblastoma. Bull Math Biol. 2020;82: 143. pmid:33159592
  39. 39. Wang CH, Rockhill JK, Mrugala M, Peacock DL, Lai A, Jusenius K, et al. Prognostic Significance of Growth Kinetics in Newly Diagnosed Glioblastomas Revealed by Combining Serial Imaging with a Novel Biomathematical Model. Cancer Research. 2009. pp. 9133–9140. pmid:19934335
  40. 40. Jackson PR, Juliano J, Hawkins-Daarud A, Rockne RC, Swanson KR. Patient-specific mathematical neuro-oncology: using a simple proliferation and invasion tumor model to inform clinical practice. Bull Math Biol. 2015;77: 846–856. pmid:25795318
  41. 41. Gaw N, Hawkins-Daarud A, Hu LS, Yoon H, Wang L, Xu Y, et al. Integration of machine learning and mechanistic models accurately predicts variation in cell density of glioblastoma using multiparametric MRI. Sci Rep. 2019;9: 10063. pmid:31296889
  42. 42. Chaddad A, Kucharczyk MJ, Daniel P, Sabri S, Jean-Claude BJ, Niazi T, et al. Radiomics in Glioblastoma: Current Status and Challenges Facing Clinical Implementation. Front Oncol. 2019;9: 374. pmid:31165039
  43. 43. Zhou M, Scott J, Chaudhury B, Hall L, Goldgof D, Yeom KW, et al. Radiomics in Brain Tumor: Image Assessment, Quantitative Feature Descriptors, and Machine-Learning Approaches. AJNR Am J Neuroradiol. 2018;39: 208–216. pmid:28982791
  44. 44. Kocher M, Ruge MI, Galldiks N, Lohmann P. Applications of radiomics and machine learning for radiotherapy of malignant brain tumors. Strahlenther Onkol. 2020;196: 856–867. pmid:32394100
  45. 45. Hu LS, Yoon H, Eschbacher JM, Baxter LC, Dueck AC, Nespodzany A, et al. Accurate Patient-Specific Machine Learning Models of Glioblastoma Invasion Using Transfer Learning. AJNR Am J Neuroradiol. 2019;40: 418–425. pmid:30819771
  46. 46. Hu LS, Ning S, Eschbacher JM, Gaw N, Dueck AC, Smith KA, et al. Multi-Parametric MRI and Texture Analysis to Visualize Spatial Histologic Heterogeneity and Tumor Extent in Glioblastoma. PLoS One. 2015;10: e0141506. pmid:26599106
  47. 47. Lu C-F, Hsu F-T, Hsieh KL-C, Kao Y-CJ, Cheng S-J, Hsu JB-K, et al. Machine Learning–Based Radiomics for Molecular Subtyping of Gliomas. Clin Cancer Res. 2018;24: 4429–4436. pmid:29789422
  48. 48. Albert FK, Forsting M, Sartor K, Adams HP, Kunze S. Early postoperative magnetic resonance imaging after resection of malignant glioma: objective evaluation of residual tumor and its influence on regrowth and prognosis. Neurosurgery. 1994;34. pmid:8121569
  49. 49. Brown TJ, Brennan MC, Li M, Church EW, Brandmeir NJ, Rakszawski KL, et al. Association of the extent of resection with survival in glioblastoma: A systematic review and meta-analysis. JAMA Oncol. 2016;2: 1460–1469. pmid:27310651
  50. 50. Berens ME, Giese A. “…those left behind.” Biology and Oncology of Invasive Glioma Cells. Neoplasia. 1999. pp. 208–219. pmid:10935475
  51. 51. Bond KM, Curtin L, Hawkins-Daarud A, Urcuyo JC, De Leon G, Sereduk C, et al. Glioblastoma states are defined by cohabitating cellular populations with progression-, imaging- and sex-distinct patterns. bioRxiv. 2022.
  52. 52. Ellingson BM, Malkin MG, Rand SD, Connelly JM, Quinsey C, LaViolette PS, et al. Validation of functional diffusion maps (fDMs) as a biomarker for human glioma cellularity. J Magn Reson Imaging. 2010;31: 538–548. pmid:20187195
  53. 53. Mekkaoui C, Metellus P, Kostis WJ, Martuzzi R, Pereira FR, Beregi J-P, et al. Diffusion Tensor Imaging in Patients with Glioblastoma Multiforme Using the Supertoroidal Model. PLoS One. 2016;11: e0146693. pmid:26761637
  54. 54. Stadlbauer A, Ganslandt O, Buslei R, Hammen T, Gruber S, Moser E, et al. Gliomas: histopathologic evaluation of changes in directionality and magnitude of water diffusion at diffusion-tensor MR imaging. Radiology. 2006;240: 803–810.
  55. 55. Barajas RF Jr, Cha S. Benefits of dynamic susceptibility-weighted contrast-enhanced perfusion MRI for glioma diagnosis and therapy. CNS Oncol. 2014;3: 407–419. pmid:25438812
  56. 56. Soni N, Priya S, Bathla G. Texture Analysis in Cerebral Gliomas: A Review of the Literature. AJNR Am J Neuroradiol. 2019;40: 928–934. pmid:31122918
  57. 57. Molina D, Pérez-Beteta J, Luque B, Arregui E, Calvo M, Borrás JM, et al. Tumour heterogeneity in glioblastoma assessed by MRI texture analysis: a potential marker of survival. Br J Radiol. 2016;89: 20160242. pmid:27319577
  58. 58. Price SJ, Jena R, Burnet NG, Hutchinson PJ, Dean AF, Peña A, et al. Improved delineation of glioma margins and regions of infiltration with the use of diffusion tensor imaging: an image-guided biopsy study. AJNR Am J Neuroradiol. 2006;27: 1969–1974. pmid:17032877
  59. 59. Brown R, Zlatescu M, Sijben A, Roldan G, Easaw J, Forsyth P, et al. The use of magnetic resonance imaging to noninvasively detect genetic signatures in oligodendroglioma. Clin Cancer Res. 2008;14: 2357–2362. pmid:18413825
  60. 60. Drabycz S, Roldán G, de Robles P, Adler D, McIntyre JB, Magliocco AM, et al. An analysis of image texture, tumor location, and MGMT promoter methylation in glioblastoma using magnetic resonance imaging. Neuroimage. 2010;49: 1398–1405. pmid:19796694
  61. 61. Weber M-A, Henze M, Tüttenberg J, Stieltjes B, Meissner M, Zimmer F, et al. Biopsy targeting gliomas: do functional imaging techniques identify similar target areas? Invest Radiol. 2010;45: 755–768. pmid:20829706
  62. 62. Hu LS, Eschbacher JM, Dueck AC, Heiserman JE, Liu S, Karis JP, et al. Correlations between perfusion MR imaging cerebral blood volume, microvessel quantification, and clinical outcome using stereotactic analysis in recurrent high-grade glioma. AJNR Am J Neuroradiol. 2012;33: 69–76. pmid:22095961
  63. 63. Hu LS, Eschbacher JM, Heiserman JE, Dueck AC, Shapiro WR, Liu S, et al. Reevaluating the imaging definition of tumor progression: perfusion MRI quantifies recurrent glioblastoma tumor fraction, pseudoprogression, and radiation necrosis to predict survival. Neuro Oncol. 2012;14: 919–930. pmid:22561797
  64. 64. Semmineh NB, Xu J, Skinner JT, Xie J, Li H, Ayers G, et al. Assessing tumor cytoarchitecture using multiecho DSC-MRI derived measures of the transverse relaxivity at tracer equilibrium (TRATE). Magn Reson Med. 2015;74: 772–784. pmid:25227668
  65. 65. Gillies RJ, Kinahan PE, Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology. 2016;278: 563–577. pmid:26579733
  66. 66. Gatenby RA, Silva AS, Gillies RJ, Roy Frieden B. Adaptive Therapy. Cancer Research. 2009. pp. 4894–4903. pmid:19487300
  67. 67. Guo Z, Shu Y, Zhou H, Zhang W, Wang H. Radiogenomics helps to achieve personalized therapy by evaluating patient responses to radiation treatment. Carcinogenesis. 2015;36: 307–317. pmid:25604391
  68. 68. Fathi Kazerooni A, Bakas S, Saligheh Rad H, Davatzikos C. Imaging signatures of glioblastoma molecular characteristics: A radiogenomics review. J Magn Reson Imaging. 2020;52: 54–69. pmid:31456318
  69. 69. Bakdash JZ, Marusich LR. Repeated Measures Correlation. Front Psychol. 2017;8: 456. pmid:28439244
  70. 70. West BT, Welch KB, Galecki AT. Linear Mixed Models: A Practical Guide Using Statistical Software. CRC Press; 2022.
  71. 71. Therneau TM, Grambsch PM. Modeling Survival Data: Extending the Cox Model. Springer Science & Business Media; 2013.
  72. 72. Outcomes KDI. KDIGO 2012 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease. 2013.
  73. 73. Institute NC, National Cancer Institute. National Human Genome Research Institute. Definitions. 2020.
  74. 74. Medtronic. Surgical Navigation Systems—StealthStation. [cited 11 Nov 2022]. https://www.medtronic.com/us-en/healthcare-professionals/products/neurological/surgical-navigation-systems/stealthstation.html
  75. 75. Southern Labware. [cited 11 Nov 2022]. www.southernlabware.com
  76. 76. IVFSTORE. In: ivfstore.com [Internet]. [cited 11 Nov 2022]. us.ivfstore.com/products/lab-line-thermo-flasks
  77. 77. Azer Scientific Pre-Filled 10% Neutral Buffered Formalin. [cited 23 May 2023]. https://www.fishersci.com/shop/products/azer-scientific-pre-filled-10-neutral-buffered-formalin-15/22026439
  78. 78. Louis DN, Perry A, Reifenberger G, von Deimling A, Figarella-Branger D, Cavenee WK, et al. The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary. Acta Neuropathol. 2016;131: 803–820. pmid:27157931
  79. 79. Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, et al. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol. 2021;23: 1231–1251. pmid:34185076
  80. 80. Bond KM, Curtin L, Ranjbar S, Afshari AE, Hu LS, Rubin JB, et al. An image-based modeling framework for predicting spatiotemporal brain cancer biology within individual patients. Frontiers in Oncology. 2023; 13:1185738. pmid:37849813