Association between patient-reported cognitive function and location of glioblastoma.

Objective cognitive function in patients with glioblastoma may depend on tumor location. Less is known about the potential impact of tumor location on cognitive function from the patients' perspective. This study aimed to investigate the association between patient-reported cognitive function and the location of glioblastoma using voxel-based lesion-symptom mapping. Patient-reported cognitive function was assessed with the European Organisation for Research and Treatment (EORTC) QLQ-C30 cognitive function subscale preoperatively and 1 month postoperatively. Semi-automatic tumor segmentations from preoperative MRI images with the corresponding EORTC QLQ-C30 cognitive function score were registered to a standardized brain template. Student's pooled-variance t-test was used to compare mean patient-reported cognitive function scores between those with and without tumors in each voxel. Both preoperative brain maps (n = 162) and postoperative maps of changes (n = 99) were developed. Glioblastomas around the superior part of the left lateral ventricle, the left lateral part of the thalamus, the left caudate nucleus, and a portion of the left internal capsule were significantly associated with reduced preoperative patient-reported cognitive function. However, no voxels were significantly associated with postoperative change in patient-reported cognitive function assessed 1 month postoperatively. There seems to be an anatomical relation between tumor location and patient-reported cognitive function before surgery, with the left hemisphere being the dominant from the patients' perspective.

How well do neurosurgeons predict survival in patients with high-grade glioma?

Due to the lack of reliable prognostic tools, prognostication and surgical decisions largely rely on the neurosurgeons' clinical prediction skills. The aim of this study was to assess the accuracy of neurosurgeons' prediction of survival in patients with high-grade glioma and explore factors possibly associated with accurate predictions. In a prospective single-center study, 199 patients who underwent surgery for high-grade glioma were included. After surgery, the operating surgeon predicted the patient's survival using an ordinal prediction scale. A survival curve was used to visualize actual survival in groups based on this scale, and the accuracy of clinical prediction was assessed by comparing predicted and actual survival. To investigate factors possibly associated with accurate estimation, a binary logistic regression analysis was performed. The surgeons were able to differentiate between patients with different lengths of survival, and median survival fell within the predicted range in all groups with predicted survival < 24 months. In the group with predicted survival > 24 months, median survival was shorter than predicted. The overall accuracy of surgeons' survival estimates was 41%, and over- and underestimations were done in 34% and 26%, respectively. Consultants were 3.4 times more likely to accurately predict survival compared to residents (p = 0.006). Our findings demonstrate that although especially experienced neurosurgeons have rather good predictive abilities when estimating survival in patients with high-grade glioma on the group level, they often miss on the individual level. Future prognostic tools should aim to beat the presented clinical prediction skills.

Are there predilection sites for intracranial meningioma? A population-based atlas.

Meningioma is the most common benign intracranial tumor and is believed to arise from arachnoid cap cells of arachnoid granulations. We sought to develop a population-based atlas from pre-treatment MRIs to explore the distribution of intracranial meningiomas and to explore risk factors for development of intracranial meningiomas in different locations. All adults (≥ 18 years old) diagnosed with intracranial meningiomas and referred to the department of neurosurgery from a defined catchment region between 2006 and 2015 were eligible for inclusion. Pre-treatment T1 contrast-enhanced MRI-weighted brain scans were used for semi-automated tumor segmentation to develop the meningioma atlas. Patient variables used in the statistical analyses included age, gender, tumor locations, WHO grade and tumor volume. A total of 602 patients with intracranial meningiomas were identified for the development of the brain tumor atlas from a wide and defined catchment region. The spatial distribution of meningioma within the brain is not uniform, and there were more tumors in the frontal region, especially parasagittally, along the anterior part of the falx, and on the skull base of the frontal and middle cranial fossa. More than 2/3 meningioma patients were females (p < 0.001) who also were more likely to have multiple meningiomas (p < 0.01), while men more often have supratentorial meningiomas (p < 0.01). Tumor location was not associated with age or WHO grade. The distribution of meningioma exhibits an anterior to posterior gradient in the brain. Distribution of meningiomas in the general population is not dependent on histopathological WHO grade, but may be gender-related.

What is the current clinico-radiological diagnostic accuracy for intracranial tumours?

OBJECTIVE: To determine the diagnostic accuracy of routine clinico-radiological workup for a population-based selection of intracranial tumours. METHODS: In this prospective cohort study, we included consecutive adult patients who underwent a primary surgical intervention for a suspected intracranial tumour between 2015 and 2019 at a single-neurosurgical centre. The treating team estimated the expected diagnosis prior to surgery using predefined groups. The expected diagnosis was compared to final histopathology and the accuracy of preoperative clinico-radiological diagnosis (sensitivity, specificity, positive and negative predictive values) was calculated. RESULTS: 392 patients were included in the data analysis, of whom 319 underwent a primary surgical resection and 73 were operated with a diagnostic biopsy only. The diagnostic accuracy varied between different tumour types. The overall sensitivity, specificity and diagnostic mismatch rate of clinico-radiological diagnosis was 85.8%, 97.7% and 4.0%, respectively. For gliomas (including differentiation between low-grade and high-grade gliomas), the same diagnostic accuracy measures were found to be 82.2%, 97.2% and 5.6%, respectively. The most common diagnostic mismatch was between low-grade gliomas, high-grade gliomas and metastases. Accuracy of 90.2% was achieved for differentiation between diffuse low-grade gliomas and high-grade gliomas. CONCLUSIONS: The current accuracy of a preoperative clinico-radiological diagnosis of brain tumours is high. Future non-invasive diagnostic methods need to outperform our results in order to add much value in a routine clinical setting in unselected patients.

Brain infarctions after glioma surgery: prevalence, radiological characteristics and risk factors.

BACKGROUND: Prevalence, radiological characteristics, and risk factors for peritumoral infarctions after glioma surgery are not much studied. In this study, we assessed shape, volume, and prevalence of peritumoral infarctions and investigated possible associated factors. METHODS: In a prospective single-center cohort study, we included all adult patients operated for diffuse gliomas from January 2007 to December 2018. Postoperative infarctions were segmented using early postoperative MRI images, and volume, shape, and location of postoperative infarctions were assessed. Heatmaps of the distribution of tumors and infarctions were created. RESULTS: MRIs from 238 (44%) of 539 operations showed restricted diffusion in relation to the operation cavity, interpreted as postoperative infarctions. Of these, 86 (36%) were rim-shaped, 103 (43%) were sector-shaped, 40 (17%) were a combination of rim- and sector-shaped, and six (3%) were remote infarctions. Median infarction volume was 1.7 cm3 (IQR 0.7-4.3, range 0.1-67.1). Infarctions were more common if the tumor was in the temporal lobe, and the map shows more infarctions in the periventricular watershed areas. Sector-shaped infarctions were more often seen in patients with known cerebrovascular disease (47.6% vs. 25.5%, p = 0.024). There was a positive correlation between infarction volume and tumor volume (r = 0.267, p < 0.001) and infarction volume and perioperative bleeding (r = 0.176, p = 0.014). Moreover, there was a significant positive association between age and larger infarction volumes (r = 0.193, p = 0.003). Infarction rates and infarction volumes varied across individual surgeons, p = 0.037 (range 32-72%) and p = 0.026. CONCLUSIONS: In the present study, peritumoral infarctions occurred in 44% after diffuse glioma operations. Infarctions were more common in patients operated for tumors in the temporal lobe but were not more common following recurrent surgeries. Sector-shaped infarctions were more common in patients with known cerebrovascular disease. Increasing age, larger tumors, and more perioperative bleeding were factors associated with infarction volumes. The risk of infarctions and infarction volumes may also be surgeon-dependent.

Survival of glioblastoma in relation to tumor location: a statistical tumor atlas of a population-based cohort.

PURPOSE: Previous studies on the effect of tumor location on overall survival in glioblastoma have found conflicting results. Based on statistical maps, we sought to explore the effect of tumor location on overall survival in a population-based cohort of patients with glioblastoma and IDH wild-type astrocytoma WHO grade II-III with radiological necrosis. METHODS: Patients were divided into three groups based on overall survival: < 6 months, 6-24 months, and > 24 months. Statistical maps exploring differences in tumor location between these three groups were calculated from pre-treatment magnetic resonance imaging scans. Based on the results, multivariable Cox regression analyses were performed to explore the possible independent effect of centrally located tumors compared to known prognostic factors by use of distance from center of the third ventricle to contrast-enhancing tumor border in centimeters as a continuous variable. RESULTS: A total of 215 patients were included in the statistical maps. Central tumor location (corpus callosum, basal ganglia) was associated with overall survival < 6 months. There was also a reduced overall survival in patients with tumors in the left temporal lobe pole. Tumors in the dorsomedial right temporal lobe and the white matter region involving the left anterior paracentral gyrus/dorsal supplementary motor area/medial precentral gyrus were associated with overall survival > 24 months. Increased distance from center of the third ventricle to contrast-enhancing tumor border was a positive prognostic factor for survival in elderly patients, but less so in younger patients. CONCLUSIONS: Central tumor location was associated with worse prognosis. Distance from center of the third ventricle to contrast-enhancing tumor border may be a pragmatic prognostic factor in elderly patients.

Spatial distribution of malignant transformation in patients with low-grade glioma.

BACKGROUND: Malignant transformation represents the natural evolution of diffuse low-grade gliomas (LGG). This is a catastrophic event, causing neurocognitive symptoms, intensified treatment and premature death. However, little is known concerning the spatial distribution of malignant transformation in patients with LGG. MATERIALS AND METHODS: Patients histopathological diagnosed with LGG and subsequent radiological malignant transformation were identified from two different institutions. We evaluated the spatial distribution of malignant transformation with (1) visual inspection and (2) segmentations of longitudinal tumor volumes. In (1) a radiological transformation site < 2 cm from the tumor on preceding MRI was defined local transformation. In (2) overlap with pretreatment volume after importation into a common space was defined as local transformation. With a centroid model we explored if there were particular patterns of transformations within relevant subgroups. RESULTS: We included 43 patients in the clinical evaluation, and 36 patients had MRIs scans available for longitudinal segmentations. Prior to malignant transformation, residual radiological tumor volumes were > 10 ml in 93% of patients. The transformation site was considered local in 91% of patients by clinical assessment. Patients treated with radiotherapy prior to transformation had somewhat lower rate of local transformations (83%). Based upon the segmentations, the transformation was local in 92%. We did not observe any particular pattern of transformations in examined molecular subgroups. CONCLUSION: Malignant transformation occurs locally and within the T2w hyperintensities in most patients. Although LGG is an infiltrating disease, this data conceptually strengthens the role of loco-regional treatments in patients with LGG.

Is the anatomical distribution of low-grade gliomas linked to regions of gliogenesis?

INTRODUCTION: According to the stem cell theory, two neurogenic niches in the adult human brain may harbor cells that initiate the formation of gliomas: The larger subventricular zone (SVZ) and the subgranular zone (SGZ) in the hippocampus. We wanted to explore whether defining molecular markers in low-grade gliomas (LGG; WHO grade II) are related to distance to the neurogenic niches. METHODS: Patients treated at two Norwegian university hospitals with population-based referral were included. Eligible patients had histopathological verified supratentorial low-grade glioma. IDH mutational status and 1p19q co-deletion status was retrospectively assessed. 159 patients were included, and semi-automatic tumor segmentation was done from pre-treatment T2-weighted (T2W) or Fluid-Attenuated Inversion Recovery (FLAIR) images. 3D maps showing the anatomical distribution of the tumors were then created for each of the three molecular subtypes (IDH mutated/1p19q co-deleted, IDH mutated and IDH wild-type). Both distance from tumor center and tumor border to the neurogenic niches were recorded. RESULTS: In this population-based cohort of previously untreated low-grade gliomas, we found that low-grade gliomas are more often found closer to the SVZ than the SGZ, but IDH wild-type tumors are more often found near SGZ. CONCLUSION: Our study suggests that the stem cell origin of IDH wild-type and IDH mutated low-grade gliomas may be different.

Intra-rater variability in low-grade glioma segmentation.

Assessment of size and growth are key radiological factors in low-grade gliomas (LGGs), both for prognostication and treatment evaluation, but the reliability of LGG-segmentation is scarcely studied. With a diffuse and invasive growth pattern, usually without contrast enhancement, these tumors can be difficult to delineate. The aim of this study was to investigate the intra-observer variability in LGG-segmentation for a radiologist without prior segmentation experience. Pre-operative 3D FLAIR images of 23 LGGs were segmented three times in the software 3D Slicer. Tumor volumes were calculated, together with the absolute and relative difference between the segmentations. To quantify the intra-rater variability, we used the Jaccard coefficient comparing both two (J2) and three (J3) segmentations as well as the Hausdorff Distance (HD). The variability measured with J2 improved significantly between the two last segmentations compared to the two first, going from 0.87 to 0.90 (p = 0.04). Between the last two segmentations, larger tumors showed a tendency towards smaller relative volume difference (p = 0.07), while tumors with well-defined borders had significantly less variability measured with both J2 (p = 0.04) and HD (p < 0.01). We found no significant relationship between variability and histological sub-types or Apparent Diffusion Coefficients (ADC). We found that the intra-rater variability can be considerable in serial LGG-segmentation, but the variability seems to decrease with experience and higher grade of border conspicuity. Our findings highlight that some criteria defining tumor borders and progression in 3D volumetric segmentation is needed, if moving from 2D to 3D assessment of size and growth of LGGs.

Intra-operative correction of brain-shift.

BACKGROUND: Brain-shift is a major source of error in neuronavigation systems based on pre-operative images. In this paper, we present intra-operative correction of brain-shift using 3D ultrasound. METHODS: The method is based on image registration of vessels extracted from pre-operative MRA and intra-operative power Doppler-based ultrasound and is fully integrated in the neuronavigation software. RESULTS: We have performed correction of brain-shift in the operating room during surgery and provided the surgeon with updated information. Here, we present data from seven clinical cases with qualitative and quantitative error measures. CONCLUSION: The registration algorithm is fast enough to provide the surgeon with updated information within minutes and accounts for large portions of the experienced shift. Correction of brain-shift can make pre-operative data like fMRI and DTI reliable for a longer period of time and increase the usefulness of the MR data as a supplement to intra-operative 3D ultrasound in terms of overview and interpretation.

Liver deformation in an animal model due to pneumoperitoneum assessed by a vessel-based deformable registration.

PURPOSE: Surgical navigation based on preoperative images partly overcomes some of the drawbacks of minimally invasive interventions - reduction of free sight, lack of dexterity and tactile feedback. The usefulness of preoperative images is limited in laparoscopic liver surgery, as the liver shifts due to respiration, induction of pneumoperitoneum and surgical manipulation. In this study, we evaluated the shift and deformation in an animal liver caused by respiration and pneumopertioneum using intraoperative cone beam CT. MATERIAL AND METHODS: 3D cone beam CT scans were acquired with arterial contrast. The centerlines of the segmented vessels were extracted from the images taken at different respiration and pressure settings. A non-rigid registration method was used to measure the shift and deformation. The mean Euclidean distance between the annotated landmarks was used for evaluation. RESULTS: A shift and deformation of 44.6 mm on average was introduced due to the combined effect of respiration and pneumoperitoneum. On average 91% of the deformations caused by the respiration and pneumoperitoneum were recovered. CONCLUSION: The results can contribute to the use of intraoperative imaging to correct for anatomic shift so that preoperative data can be used with greater confidence and accuracy during guidance of laparoscopic liver procedures.

Incidence, risk factors, and clinical implications of postoperative blood in or near the resection cavity after glioma surgery.

Introduction: Postoperative hematomas that require reoperation are a serious, but uncommon complication to glioma surgery. However, smaller blood volumes are frequently observed, but their clinical significance is less known. Research question: What are the incidence rates, risk factors, and patient-reported outcomes of all measurable blood in or near the resection cavity on postoperative MRI in diffuse glioma patients? Material and methods: We manually segmented intradural and extradural blood from early postoperative MRI of 292 diffuse glioma resections. Potential associations between blood volume and tumor characteristics, demographics, and perioperative factors were explored using non-parametric methods. The assessed outcomes were generic and disease-specific patient-reported HRQoL. Results: Out of the 292 MRI scans included, 184 (63%) had intradural blood, and 212 (73%) had extradural blood in or near the resection cavity. The median blood volumes were 0.4 mL and 3.0 mL, respectively. Intradural blood volume was associated with tumor volume, intraoperative blood loss, and EOR. Extradural blood volume was associated with age and tumor volume. Greater intradural blood volume was associated with less headache and cognitive improvement, but not after adjustments for tumor volume. Discussion and conclusions: Postoperative blood on early postoperative MRI is common. Intradural blood volumes tend to be larger in patients with larger tumors, more intraoperative blood loss, or undergoing subtotal resections. Extradural blood volumes tend to be larger in younger patients with larger tumors. Postoperative blood in or near the resection cavity that does not require reoperation does not seem to affect HRQoL in diffuse glioma patients.

Lesion Frequency Distribution Maps of Traumatic Axonal Injury on Early Magnetic Resonance Imaging After Moderate and Severe Traumatic Brain Injury and Associations to 12 Months Outcome.

Traumatic axonal injury (TAI) is a common finding on magnetic resonance imaging (MRI) in patients with moderate-severe traumatic brain injury (TBI), and the burden of TAI is associated with outcome in this patient group. Lesion mapping offers a way to combine imaging findings from numerous individual patients into common lesion maps where the findings from a whole patient cohort can be assessed. The aim of this study was to evaluate the spatial distribution of TAI lesions on different MRI sequences and its associations to outcome with use of lesion mapping. Included prospectively were 269 patients (8-70 years) with moderate or severe TBI and MRI within six weeks after injury. The TAI lesions were evaluated and manually segmented on fluid-attenuated inversed recovery (FLAIR), diffusion weighted imaging (DWI), and either T2* gradient echo (T2*GRE) or susceptibility weighted imaging (SWI). The segmentations were registered to the Montreal Neurological Institute space and combined to lesion frequency distribution maps. Outcome was assessed with Glasgow Outcome Scale Extended (GOSE) score at 12 months. The frequency and distribution of TAI was assessed qualitatively by visual reading. Univariable associations to outcome were assessed qualitatively by visual reading and also quantitatively with use of voxel-based lesion-symptom mapping (VLSM). The highest frequency of TAI was found in the posterior half of corpus callosum. The frequency of TAI was higher in the frontal and temporal lobes than in the parietal and occipital lobes, and in the upper parts of the brainstem than in the lower. At the group level, all voxels in mesencephalon had TAI on FLAIR. The patients with poorest outcome (GOSE scores ≤4) had higher frequencies of TAI. On VLSM, poor outcome was associated with TAI lesions bilaterally in the splenium, the right side of tectum, tegmental mesencephalon, and pons. In conclusion, we found higher frequency of TAI in posterior corpus callosum, and TAI in splenium, mesencephalon, and pons were associated with poor outcome. If lesion frequency distribution maps containing outcome information based on imaging findings from numerous patients in the future can be compared with the imaging findings from individual patients, it would offer a new tool in the clinical workup and outcome prediction of the patient with TBI.

Growth dynamics of untreated meningiomas.

Background: Knowledge about meningioma growth characteristics is needed for developing biologically rational follow-up routines. In this study of untreated meningiomas followed with repeated magnetic resonance imaging (MRI) scans, we studied growth dynamics and explored potential factors associated with tumor growth. Methods: In a single-center cohort study, we included 235 adult patients with radiologically suspected intracranial meningioma and at least 3 MRI scans during follow-up. Tumors were segmented using an automatic algorithm from contrast-enhanced T1 series, and, if needed, manually corrected. Potential meningioma growth curves were statistically compared: linear, exponential, linear radial, or Gompertzian. Factors associated with growth were explored. Results: In 235 patients, 1394 MRI scans were carried out in the median 5-year observational period. Of the models tested, a Gompertzian growth curve best described growth dynamics of meningiomas on group level. 59% of the tumors grew, 27% remained stable, and 14% shrunk. Only 13 patients (5%) underwent surgery during the observational period and were excluded after surgery. Tumor size at the time of diagnosis, multifocality, and length of follow-up were associated with tumor growth, whereas age, sex, presence of peritumoral edema, and hyperintense T2-signal were not significant factors. Conclusions: Untreated meningiomas follow a Gompertzian growth curve, indicating that increasing and potentially doubling subsequent follow-up intervals between MRIs seems biologically reasonable, instead of fixed time intervals. Tumor size at diagnosis is the strongest predictor of future growth, indicating a potential for longer follow-up intervals for smaller tumors. Although most untreated meningiomas grow, few require surgery.

Standardized evaluation of the extent of resection in glioblastoma with automated early post-operative segmentation.

Standard treatment of patients with glioblastoma includes surgical resection of the tumor. The extent of resection (EOR) achieved during surgery significantly impacts prognosis and is used to stratify patients in clinical trials. In this study, we developed a U-Net-based deep-learning model to segment contrast-enhancing tumor on post-operative MRI exams taken within 72 h of resection surgery and used these segmentations to classify the EOR as either maximal or submaximal. The model was trained on 122 multiparametric MRI scans from our institution and achieved a mean Dice score of 0.52 ± 0.03 on an external dataset (n = 248), a performance -on par with the interrater agreement between expert annotators as reported in literature. We obtained an EOR classification precision/recall of 0.72/0.78 on the internal test dataset (n = 462) and 0.90/0.87 on the external dataset. Furthermore, Kaplan-Meier curves were used to compare the overall survival between patients with maximal and submaximal resection in the internal test dataset, as determined by either clinicians or the model. There was no significant difference between the survival predictions using the model's and clinical EOR classification. We find that the proposed segmentation model is capable of reliably classifying the EOR of glioblastoma tumors on early post-operative MRI scans. Moreover, we show that stratification of patients based on the model's predictions offers at least the same prognostic value as when done by clinicians.

Ultrasound imaging in neurosurgery: approaches to minimize surgically induced image artefacts for improved resection control.

BACKGROUND: Intraoperative ultrasound imaging is used in brain tumor surgery to identify tumor remnants. The ultrasound images may in some cases be more difficult to interpret in the later stages of the operation than in the beginning of the operation. The aim of this paper is to explain the causes of surgically induced ultrasound artefacts and how they can be recognized and reduced. METHODS: The theoretical reasons for artefacts are addressed and the impact of surgery is discussed. Different setups for ultrasound acquisition and different acoustic coupling fluids to fill up the resection cavity are evaluated with respect to improved image quality. RESULTS: The enhancement artefact caused by differences in attenuation of the resection cavity fluid and the surrounding brain is the most dominating surgically induced ultrasound artefact. The influence of the artefact may be reduced by inserting ultrasound probes with small footprint into the resection cavity for a close-up view of the areas with suspected tumor remnants. A novel acoustic coupling fluid developed for use during ultrasound imaging in brain tumor surgery has the potential to reduce surgically induced ultrasound artefacts to a minimum. CONCLUSIONS: Surgeons should be aware of artefacts in ultrasound images that may occur during brain tumor surgery. Techniques to identify and reduce image artefacts are useful and should be known to users of ultrasound in brain tumor surgery.

The Big Bang of Deep Learning in Ultrasound-Guided Surgery: A Review.

Ultrasound (US) imaging is a paramount modality in many image-guided surgeries and percutaneous interventions, thanks to its high portability, temporal resolution, and cost-efficiency. However, due to its imaging principles, the US is often noisy and difficult to interpret. Appropriate image processing can greatly enhance the applicability of the imaging modality in clinical practice. Compared with the classic iterative optimization and machine learning (ML) approach, deep learning (DL) algorithms have shown great performance in terms of accuracy and efficiency for US processing. In this work, we conduct a comprehensive review on deep-learning algorithms in the applications of US-guided interventions, summarize the current trends, and suggest future directions on the topic.

Raidionics: an open software for pre- and postoperative central nervous system tumor segmentation and standardized reporting.

For patients suffering from central nervous system tumors, prognosis estimation, treatment decisions, and postoperative assessments are made from the analysis of a set of magnetic resonance (MR) scans. Currently, the lack of open tools for standardized and automatic tumor segmentation and generation of clinical reports, incorporating relevant tumor characteristics, leads to potential risks from inherent decisions' subjectivity. To tackle this problem, the proposed Raidionics open-source software has been developed, offering both a user-friendly graphical user interface and stable processing backend. The software includes preoperative segmentation models for each of the most common tumor types (i.e., glioblastomas, lower grade gliomas, meningiomas, and metastases), together with one early postoperative glioblastoma segmentation model. Preoperative segmentation performances were quite homogeneous across the four different brain tumor types, with an average Dice around 85% and patient-wise recall and precision around 95%. Postoperatively, performances were lower with an average Dice of 41%. Overall, the generation of a standardized clinical report, including the tumor segmentation and features computation, requires about ten minutes on a regular laptop. The proposed Raidionics software is the first open solution enabling an easy use of state-of-the-art segmentation models for all major tumor types, including preoperative and postsurgical standardized reports.

Segmentation of glioblastomas in early post-operative multi-modal MRI with deep neural networks.

Extent of resection after surgery is one of the main prognostic factors for patients diagnosed with glioblastoma. To achieve this, accurate segmentation and classification of residual tumor from post-operative MR images is essential. The current standard method for estimating it is subject to high inter- and intra-rater variability, and an automated method for segmentation of residual tumor in early post-operative MRI could lead to a more accurate estimation of extent of resection. In this study, two state-of-the-art neural network architectures for pre-operative segmentation were trained for the task. The models were extensively validated on a multicenter dataset with nearly 1000 patients, from 12 hospitals in Europe and the United States. The best performance achieved was a 61% Dice score, and the best classification performance was about 80% balanced accuracy, with a demonstrated ability to generalize across hospitals. In addition, the segmentation performance of the best models was on par with human expert raters. The predicted segmentations can be used to accurately classify the patients into those with residual tumor, and those with gross total resection.