Background Glioblastoma is the commonest malignant brain tumour. Sarcopenia is associated with worse cancer survival, but manually quantifying muscle on imaging is time-consuming. We present a deep learning-based system for quantification of temporalis muscle, a surrogate for skeletal muscle mass, and assess its prognostic value in glioblastoma. Methods A neural network for temporalis segmentation was trained with 366 MRI head images from 132 patients from 4 different glioblastoma data sets and used to quantify muscle cross-sectional area (CSA). Association between temporalis CSA and survival was determined in 96 glioblastoma patients from internal and external data sets. Results The model achieved high segmentation accuracy (Dice coefficient 0.893). Median age was 55 and 58 years and 75.6 and 64.7% were males in the in-house and TCGA-GBM data sets, respectively. CSA was an independently significant predictor for survival in both the in-house and TCGA-GBM data sets (HR 0.464, 95% CI 0.218–0.988, p = 0.046; HR 0.466, 95% CI 0.235–0.925, p = 0.029, respectively). Conclusions Temporalis CSA is a prognostic marker in patients with glioblastoma, rapidly and accurately assessable with deep learning. We are the first to show that a head/neck muscle-derived sarcopenia metric generated using deep learning is associated with oncological outcomes and one of the first to show deep learning-based muscle quantification has prognostic value in cancer.

No abstract

AIMS The Gliocova dataset uses linked English national cancer data on all 51,775 adult primary brain tumour patients diagnosed between 2013-2018. Here we investigate patient safety and post-operative complications after first surgical intervention. METHOD We identified patients undergoing first surgical intervention (surgical debulking or biopsy) and used a modified Delphi approach to identify diagnosis codes indicating potential post-surgical complications. We calculated Elixhauer Comorbidity Index (ECI) weights based on our data and developed regression models to link patient characteristics and ECI with 30-day mortality, readmission and chance of complication. RESULTS 29,258 out of 51,775 patients underwent a surgical intervention (28,173 surgical debulking; and 1,207 biopsy). 11,959 (40.9\%) patients had at least one comorbidity during first intervention admission. In hospital mortality was 0.99\% (N = 289), 30-day mortality was 2.3\% (N = 677) and 30-day readmission was 12.7\% (N = 3,725). 13,137 patients (44.9\%) had at least one complication code from our defined list, either during their first surgical intervention or during a 30-day readmission. Predictors of 30-day mortality, readmission, and risk of complications included age, ECI score, number of complications, type of intervention (biopsy vs surgical debulking), income quintile, and tumour type (i.e., Glioblastoma versus other types of brain tumours). CONCLUSION To our knowledge this is the first study in England to assess post-surgical complications in a large brain tumour patient cohort. Our further work will focus on variation in outcomes between different centres/ centre volumes/ regions and the cost of complications. More information: https://blogs.imperial.ac.uk/gliocova/about-gliocova/.

AIMS The Gliocova dataset uses linked English national cancer data on all 51,775 adult primary brain tumour patients diagnosed between 2013-2018. Here we present detailed analysis of first-line treatments of adult glioblastoma (GBM) patients. METHOD We identified all adults patients diagnosed with a GBM. We focused on the first line of treatment and we defined ‘maximal’ first-line treatment as surgical resection followed by chemo-radiotherapy with 59-60 Gy and with at least one cycle of adjuvant chemotherapy Temozolomide. RESULTS 15,294 patients were diagnosed with a glioblastoma (60% male) with a median age of 66. 79% of patients received some treatment, with younger patients more likely to be treated (>90%, 18 - 59; < 30%, > 80). 54% underwent debulking surgery; 23%, biopsy. 14% received ‘maximal’ treatment and 21%, none. Patients who had no treatment had a median survival of 2 months whereas patients who received ‘maximal’ treatment had a median survival of 16 months. CONCLUSION Most adult patients with a GBM in England have a histological diagnosis, and some oncological treatment. However, only 14% receive ‘maximal’ treatment. Of the 3222 patients who received none, some of these may have had purely private treatment; however, our dataset includes any private sector work undertaken in NHS hospitals. Survival remains poor, but outcomes in those receiving maximal treatment match those from clinical trials. However, most patients do not receive maximal treatment, and so the easiest route to improving outcomes may be optimise delivery of treatment in the 65% of patients who receive sub-maximal treatment. More information on https://blogs.imperial.ac.uk/gliocova

Aims Glioblastoma multiforme (GBM) is an aggressive brain malignancy. Performance status is an important prognostic factor but is subjectively evaluated, resulting in inaccuracy. Objective markers of frailty/physical condition, such as measures of skeletal muscle mass can be evaluated on cross-sectional imaging and is associated with cancer survival. In GBM, temporalis muscle has been identified as a skeletal muscle mass surrogate and a prognostic factor. However, current manual muscle quantification is time consuming, limiting clinical adoption. We previously developed a deep learning system for automated temporalis muscle quantification, with high accuracy (Dice coefficient 0.912), and showed muscle cross-sectional area is independently significantly associated with survival in GBM (HR 0.380). However, it required manual selection of the temporalis muscle-containing MRI slice. Thus, in this work we aimed to develop a fully automatic deep-learning system, using the eyeball as an anatomic landmark for automatic slice selection, to quantify temporalis and validate on independent datasets. Method 3D brain MRI scans were obtained from four datasets: our in-house glioblastoma patient dataset, TCGA-GBM, IVY-GAP and REMBRANDT. Manual eyeball and temporalis segmentations were performed on 2D MRI images by two experienced readers. Two neural networks (2D U-Nets) were trained, one to automatically segment the eyeball and the other to segment the temporalis muscle on 2D MRI images using Dice loss function. The cross sectional area of eyeball segmentations were quantified and thresholded, to select the superior orbital MRI slice from each scan. This slice underwent temporalis segmentation, whose cross sectional area was then quantified. Accuracy of automatically predicted eyeball and temporalis segmentations were compared to manual ground truth segmentations on metrics of Dice coefficient, precision, recall and Hausdorff distance. Accuracy of MRI slice selection (by the eyeball segmentation model) for temporalis segmentation was determined by comparing automatically selected slices to slices selected manually by a trained neuro-oncologist. Results 398 images from 185 patients and 366 images from 145 patients were used for the eyeball and temporalis segmentation models, respectively. 61 independent TCGA-GBM scans formed a validation cohort to assess the performance of the full pipeline. The model achieved high accuracy in eyeball segmentation, with test set Dice coefficient of 0.9029 ± 0.0894, precision of 0.8842 ± 0.0992, recall of 0.9297 ± 0.6020 and Hausdorff distance of 2.8847 ± 0.6020. High segmentation accuracy was also achieved by the temporalis segmentation model, with Dice coefficient of 0.8968 ± 0.0375, precision of 0.8877 ± 0.0679, recall of 0.9118 ± 0.0505 and Hausdorff distance of 1.8232 ± 0.3263 in the test set. 96.1% of automatically selected slices for temporalis segmentation were within 2 slices of the manually selected slice. Conclusion Temporalis muscle cross-sectional area can be rapidly and accurately assessed from 3D MRI brain scans using a deep learning-based system in a fully automated pipeline. Combined with our and others’ previous results that demonstrate the prognostic significance of temporalis cross-sectional area and muscle width, our findings suggest a role for deep learning in muscle mass and sarcopenia screening in GBM, with the potential to add significant value to routine imaging. Possible clinical applications include risk profiling, treatment stratification and informing interventions for muscle preservation. Further work will be to validate the prognostic value of temporalis muscle cross sectional area measurements generated by our fully automatic deep learning system in the multiple in-house and external datasets.

Background Primary brain tumours are rare, but are the cause of the most life-years lost of any cancer. Given their poor prognosis, questions about their treatment and care costs are important especially in publicly-funded healthcare systems. There is currently very little robust data on the cost of care for brain tumour patients. Here we present up-to-date estimations of the direct inpatient care costs of all adult primary brain tumour patients in England, with a specific breakdown of costs attributed to cranial glioblastomas and cranial meningiomas, the commonest primary brain tumours. Material and Methods GlioCova uses a linked English national cancer data on over 50,000 adult primary brain tumour patients diagnosed between 2013-2018, with data on secondary healthcare activities three months before diagnosis - the last three months of 2012, and follow-up data after diagnosis - up to the last admission of 2019. We examined inpatient care with a breakdown of costs attributed to different treatment types: neurosurgery, chemotherapy and radiotherapy. We used the NHS HRG4+ Reference Costs Grouper 2017/2018 to assign costs matched from the National Schedule of Reference Costs 2017/2018. We converted all values to 2021 using the Health CPI Index. Results There were a total of 51,775 adult primary brain tumour patients diagnosed in England in the 6 year period between 2013 and 2018. 48,608 of these were admitted to hospital between the last three months of 2012 and the end of 2019, of which we were able to assign costs to 47,521 patients (98%). Total inpatient costs for the whole brain tumour cohort were over £973 million during the time period (34% attributed to inpatient care for neurosurgery, chemotherapy, and radiotherapy). 14,691 (31%) of patients were diagnosed with a cranial glioblastoma, and 9,501 (20%) of patients were diagnosed with a cranial meningioma. Total inpatient costs were £349 million and £163 million for glioblastoma and meningioma respectively, with 34% again devoted to direct treatment costs. Conclusion The estimated direct inpatient care costs for all 51,775 primary brain tumour patients diagnosed in England between 2013 and 2018 were nearly £1 billion (£166 million/year cohort), 2/3 of which was not for direct treatment. This does not include outpatient chemotherapy and radiotherapy costs, other outpatient appointments, patient out-of-pocket costs, primary care, social care, or end-of-life care costs. Future work will examine variation in care and costs and extend the analysis to include outpatients. We hope these data will help make an economic argument for improving care for brain tumour patients. More information on GlioCova: https://blogs.imperial.ac.uk/gliocova/about-gliocova/.

Background The Gliocova dataset uses linked English national cancer data on all 51,775 adult primary brain tumour patients diagnosed between 2013-2018. Here we are investigating the effect of surgeon and centre volume on post-surgical outcomes such as 30-day mortality, readmission and risk of complication. Material and Methods We selected patients from our dataset that have undergone surgical debulking or biopsy for the first time. We calculated surgeon and centre volume per 6 year period. To exclude non-tumour surgeons who might do occasional tumour cases as an emergency, we removed those that performed less than 12 operations over the 6 year period and to remove newly qualified/ near retirement surgeons we removed those that did not perform at least 1 operation within the 6 first months of 2013 and last 6 months of 2018. We developed univariate regression models to link surgeon and centre volume with 30-day mortality, readmission and risk of complication and are currently working on multivariate models including other predictors such as age, sex, Elixhauer comorbidity index, deprivation status, type of intervention (biopsy vs surgical debulking), and tumour type (i.e., Glioblastoma versus other types of brain tumours). Results We identified 29,258 patients (out of 51,775) that have undergone a surgical intervention (28,173 surgical debulking; and 1,207 biopsy). In-hospital mortality was 0.99% (N = 289), 30-day mortality was 2.3% (N = 677) and 30-day readmission was 12.7% (N = 3,725). No surgeons performed less than 1 operation during the first 6 months of 2013 or last 6 months of 2018. In univariate analyses, centre volume predicted the risk of 30-day mortality (p-value<0.01) and risk of complication (p-value<0.001), whereas surgeon volume only significantly predicted risk of complication (p-value<0.001). Neither surgeon volume nor centre volume were statistically significant predictors of 30-day readmission in univariate analysis. Conclusion As most surgeon and centre volume studies in brain tumour patients come from the United States, our study is one of a few in Europe to investigate this in a large adult brain tumour population. In our future work we will refine the model to account for patient factors, and assess the cost of complications and variation of these across centres, tumour types and other patient and hospital level characteristics. More information about the Gliocova project can be accessed here: https://blogs.imperial.ac.uk/gliocova/about-gliocova/. GlioCova is supported by the Imperial/NIHR BRC, and the members of the GlioCova EAG. This work uses data provided by patients and collected by the NHS as part of their care and support

Background The Gliocova dataset uses linked English national cancer data on all 51,775 adult primary brain tumour patients diagnosed between 2013-2018. Here we present early analysis of inpatient admissions of adult glioblastoma (GBM) patients. Material and Methods We identified all adults patients diagnosed with a GBM and extracted all the inpatient admissions for 1 night or more after the date of diagnosis. We focused on number and length of admissions, variation in those numbers, and place of discharge. Results Between 2013 and 2018, 15,294 patients were diagnosed with a glioblastoma in England (60% male) with a median age of 66, of whom, 12,441 (61% male) were admitted overnight with a total of 49,384 admissions post-diagnosis. Half of these patients were less than 64 at the time of their admission. The mean number of post-diagnosis admissions was 4, with a mean length of stay of 9.5 days. However, for half of the admissions, patients stayed 5 days or less in hospital (IQR = 10). Most of the procedures done were treatment-related, such as surgery for which patients stayed an average of 6.3 days (median = 4; IQR = 5). Patients who were admitted for non-treatment reasons stayed on average almost 10 days (median = 5; IQR = 11). Fewer than 3,000 admissions resulted in a patient death (5.5% of all admissions), whereas over 23,000 admissions for 10,426 patients ended with patients being discharged at home. About 2,000 patients were discharged to another hospital, hospice or a nursing home. Conclusion Most of the patients diagnosed with a glioblastoma will be admitted at some point after their diagnosis. Although the average length of stay is not that long, there is a considerable tail of longer-staying patients, for whom improved services and support might enable quicker discharge. To our knowledge, this is the first time inpatient admissions in adult brain patients are being looked at.GlioCova is supported by the Imperial/NIHR BRC, and the members of the GlioCova EAG. This work uses data provided by patients and collected by the NHS as part of their care and support.More information on the Gliocova project can be found on https://blogs.imperial.ac.uk/gliocova.