Timothy L. Jones

Brain tumor classification using the diffusion tensor image segmentation (D-SEG) technique

Background There is an increasing demand for noninvasive brain tumor biomarkers to guide surgery and subsequent oncotherapy. We present a novel whole-brain diffusion tensor imaging (DTI) segmentation (D-SEG) to delineate tumor volumes of interest (VOIs) for subsequent classification of tumor type. D-SEG uses isotropic (p) and anisotropic (q) components of the diffusion tensor to segment regions with similar diffusion characteristics. Methods DTI scans were acquired from 95 patients with low- and high-grade glioma, metastases, and meningioma and from 29 healthy subjects. D-SEG uses k-means clustering of the 2D (p,q) space to generate segments with different isotropic and anisotropic diffusion characteristics. Results Our results are visualized using a novel RGB color scheme incorporating p, q and T2-weighted information within each segment. The volumetric contribution of each segment to gray matter, white matter, and cerebrospinal fluid spaces was used to generate healthy tissue D-SEG spectra. Tumor VOIs were extracted using a semiautomated flood-filling technique and D-SEG spectra were computed within the VOI. Classification of tumor type using D-SEG spectra was performed using support vector machines. D-SEG was computationally fast and stable and delineated regions of healthy tissue from tumor and edema. D-SEG spectra were consistent for each tumor type, with constituent diffusion characteristics potentially reflecting regional differences in tissue microstructure. Support vector machines classified tumor type with an overall accuracy of 94.7%, providing better classification than previously reported. Conclusions D-SEG presents a user-friendly, semiautomated biomarker that may provide a valuable adjunct in noninvasive brain tumor diagnosis and treatment planning.

Proposal for a prospective multi-centre audit of chronic subdural haematoma management in the United Kingdom and Ireland

Abstract Background. Chronic subdural haematoma (CSDH) is a common condition that increases in incidence with rising age. Evacuation of a CSDH is one of the commonest neurosurgical procedures; however the optimal peri-operative management, surgical technique, post-operative care and the role of adjuvant therapies remain controversial. Aim. We propose a prospective multi-centre audit in order to establish current practices, outcomes and national benchmarks for future studies. Methods. Neurosurgical units (NSU) in the United Kingdom and Ireland will be invited to enrol patients to this audit. All adult patients aged 16 years and over with a primary or recurrent CSDH will be eligible for inclusion. Outcome measures and analysis. The proposed outcome measures are (1) clinical recurrence requiring re-operation within 60 days; (2) modified Rankin scale (mRS) score at discharge from NSU; (3) morbidity and mortality in the NSU; (4) destination at discharge from NSU and (5) length of stay in the NSU. Audit standards have been derived from published systematic reviews and a recent randomised trial. The proposed standards are clinical recurrence rate < 20%; unfavourable mRS (4–6) at discharge from NSU < 30%; mortality rate in NSU < 5%; morbidity rate in NSU < 10%. Data will be submitted directly into a secure online database and analysed by the studys management group. Conclusions. The audit will determine the contemporary management and outcomes of patients with CSDH in the United Kingdom and Ireland. It will inform national guidelines, clinical practice and future studies in order to improve the outcome of patients with CSDH.

A report from the inaugural meeting of the British Neurosurgical Trainee Research Collaborative held in the Royal College of Surgeons of England, 19 October 2012.

Abstract Clinical research, which is essential for improving patient outcomes, is increasingly carried out in the context of networks established between multiple institutions. Research is also considered an important component of training curricula. The recent successful completion of a randomised trial (ROSSINI), which was led by general surgical trainees of the West Midlands Research Collaborative, has established the feasibility of trainee collaborative research networks. A research network for neurosurgical trainees in the UK and Ireland was, therefore, established following the meeting of the British Neurosurgical Trainee Association (BNTA) in Aberdeen on 19 April 2012. This BNTA initiative quickly gained the full support from the Society of British Neurological Surgeons and the UK Neurosurgical Research Network. The inaugural meeting of the British Neurosurgical Trainee Research Collaborative took place at the Royal College of Surgeons of England, London, on 19 October 2012. The purpose of this report is both to record progress to date and to promote this concept.

Discrimination between glioblastoma multiforme and solitary metastasis using morphological features derived from the p:q tensor decomposition of diffusion tensor imaging

The management and treatment of high‐grade glioblastoma multiforme (GBM) and solitary metastasis (MET) are very different and influence the prognosis and subsequent clinical outcomes. In the case of a solitary MET, diagnosis using conventional radiology can be equivocal. Currently, a definitive diagnosis is based on histopathological analysis on a biopsy sample. Here, we present a computerised decision support framework for discrimination between GBM and solitary MET using MRI, which includes: (i) a semi‐automatic segmentation method based on diffusion tensor imaging; (ii) two‐dimensional morphological feature extraction and selection; and (iii) a pattern recognition module for automated tumour classification. Ground truth was provided by histopathological analysis from pre‐treatment stereotactic biopsy or at surgical resection. Our two‐dimensional morphological analysis outperforms previous methods with high cross‐validation accuracy of 97.9% and area under the receiver operating characteristic curve of 0.975 using a neural networks‐based classifier. Copyright

Morphometric model for discrimination between glioblastoma multiforme and solitary metastasis using three-dimensional shape analysis

Glioblastoma multiforme (GBM) and brain metastasis (MET) are the most common intra‐axial brain neoplasms in adults and often pose a diagnostic dilemma using standard clinical MRI. These tumor types require different oncological and surgical management, which subsequently influence prognosis and clinical outcome.

Patient safety and image transfer between referring hospitals and neuroscience centres: could we do better?

Introduction. District general hospital scanners have historically been linked to regional neuroscience units for specialist opinions on scans and to make decisions on transfer of patients requiring neurosurgical management. The implementation of digital picture archiving and communication systems (PACS) in all hospitals in the UK has disrupted these dedicated links and technical and information governance issues have delayed reprovision of electronic transfer of images for rapid expert decision making in this group of patients. We studied improvement in image transfer to acute neurosurgery units over a 4-year period. Methods. Four-year sequential review of national provision of image transfer facilities into neurosurgery units; observational study of delays associated with image transfer modalities in one representative tertiary referral centre. Results. During the 4 years of study, all hospitals nationally have implemeted digital PACS systems for image viewing. Remote image viewing facilities have gradually changed with dedicated image links being replaced by remote PACS access. However, a minority of referrals (12%) still require images to be physically transferred between hospitals using couriers for CD-ROMs. The detailed study within our own unit shows that this adds a mean delay of 5.8 h to decision making. Conclusions. Image transfer in neuroscience has been neglected following the shift to PACS servers. The recommendations of the 2004 Neuroscience Critical Care Report are unmet and patient safety is being threatened by a continued failure to implement a coordinated solution to this problem.

Has the impact of the working time regulations changed neurosurgical trainees’ attitudes towards the European working time directive 5 years on?

Abstract We report the results from a survey of the British Neurosurgical Trainees’ Association which aimed to assess current rota patterns and their compliance with the governments working time regulations. The survey questioned whether trainees felt that shift working, imposed as a result of the European working time directive, is continuing to impact on patient care and training opportunities in neurosurgery. The responses to this survey indicate that neurosurgical trainees remain concerned with the impact that the current working time regulations have on all facets of their work: training, work– life balance, and the provision of patient care. The survey comments show that the majority would support a change in legislation to allow greater flexibility in the working time regulations.

Time to diagnosis of axial spondylarthritis in clinical practice: signs of improving awareness?

SIR, The spectrum of axial spondylarthritis (SpA) ranges from symptoms of mild inflammatory back pain to the diagnosis of AS. This group of diseases affects the vertebrae, sacroiliac joints (SIJs) and their associated entheses with a pathophysiological process ranging from acute inflammation to an exaggerated repair process resulting in excessive new bone formation. This disease has the potential to lead to structural changes, deformity and loss of function. Historically there has been a significant delay in the diagnosis of axial SpA as a result of the lack of awareness of the early stages of the disease, lack of clinical findings on examination, inconsistent imaging findings and overreliance on radiographic changes to make a diagnosis. The most recent and comprehensive study by Feldtkeller et al. [1] reports an average time to diagnosis of 10 years. Such a delay in diagnosis results in prolonged periods of pain and disease activity without a diagnosis, loss of confidence in the ability of the medical system to diagnose and treat the disease and delay in the implementation of therapies that can effectively treat the disease symptomatically and potentially prevent radiographic progression [2, 3]. The lack of diagnostic criteria facilitating the recognition of axial SpA and the poor awareness of the spectrum of disease across primary care has contributed to this wellrecognized delay. The Assessment of SpondylArthritis International Society (ASAS) classification criteria has helped with the recognition of those patients with inflammatory back pain without radiographic changes by highlighting the importance of MRI in detecting early disease and also by recognizing disease without radiographic evidence of sacroiliitis [4]. Within the military, a specific back pain pathway for the aid of primary care doctors and community physiotherapists has been in place since 2009 (Fig. 1). This promotes recognition of inflammatory back pain symptoms and axial SpA and also promotes early referral to Defence Medical Rehabilitation Centre (DMRC) Headley Court, a specialist rheumatology and rehabilitation facility for military personnel. This, in combination with the new diagnostic criteria for axial SpA, has resulted in a faster time to diagnosis of axial SpA compared with the previously published findings of Feldtkeller et al. [1]. A cohort of 138 service personnel with a diagnosis of axial SpA who attended the axial SpA inpatient rehabilitation and education course at DMRC Headley Court from March 2010 to November 2012 were reviewed as part of an assessment of the efficacy of the service in making a prompt and timely diagnosis. Information regarding age, time to diagnosis and radiological findings were retrieved from medical records. The mean age of patients was 35.2 years, 73.5% were diagnosed with AS, 20.5% with undifferentiated nonradiographic axial SpA and 6% with psoriatic spondylitis. X-ray of the SIJs was used to confirm a diagnosis of AS. MRI scanning of the SIJs or the SIJs and the spine was used in 78% of patients to assist in the diagnosis and assessment of disease. The mean time to diagnosis for all patients was 5.72 years (range 0 25) from symptom onset, with a mean age of diagnosis of 31.9 years (range 19 49) and a mean age at symptom onset of 25.8 years (range 10 48). The time to diagnosis between the sexes was quite markedly different, with SpA being diagnosed on average 3 years later in women (male 5.56 years, female 8.5 years). These data highlight a significantly reduced time to diagnosis of axial SpA/AS in this cohort of military axial SpA patients. The mean delay reported was 5.72 years, 4.3 years faster than the previously published landmark study [2]. These data were collected in a military population that is serviced by specialist outpatient clinics and inpatient rehabilitation services for axial SpA. The ASAS criteria have been fully embraced by the clinicians running the clinics and so has the role of MRI in making an early diagnosis and assessing disease activity and predicting prognosis [6, 7]. An important element of this reduced delay in Letters to the Editor

Delineation of gliomas using radial metabolite indexing

1H MRSI has demonstrated the ability to characterise and delineate brain tumours, but robust data analysis methods are still needed. In this study, we present an objective analysis method for MRSI data to delineate tumour abnormality regions. The presented method is a development of the choline‐to‐N‐acetylaspartate index (CNI), which uses perpendicular distances in a choline versus N‐acetylaspartate plot as a measure of abnormality. We propose a radial CNI (rCNI) method that uses the choline to N‐acetylaspartate ratio directly as an abnormality measure. To avoid problems with small or zero denominators, we perform an arctangent transformation. CNI abnormality contours were evaluated using a z‐score threshold of 2 (CNI2) and 2.5 (CNI2.5) and compared with rCNI2. Simulations modelling low‐grade (LGG) and high‐grade (HGG) gliomas with different tissue compartments and partial volume effects suggest improved specificity of rCNI2 (LGG 92%/HGG 91%) over CNI2 (LGG 69%/HGG 69%) and CNI2.5 (LGG 74%/HGG 75%), whilst retaining a similar sensitivity to both CNI2 and CNI2.5. Our simulation results also confirm a previously reported increase in specificity of CNI2.5 over CNI2 with little penalty in sensitivity. The analysis of MRSI data acquired from 10 patients with low‐grade glioma at 3 T suggests a more robust delineation of the lesions using rCNI with respect to conventional imaging compared with standard CNI. Further analysis of 29 glioma datasets acquired at 1.5 T, together with previously published estimated tumour proportions, suggests that rCNI has higher sensitivity and specificity for the identification of abnormal MRSI voxels. Copyright

Supervised learning based multimodal MRI brain tumour segmentation using texture features from supervoxels

BACKGROUND Accurate segmentation of brain tumour in magnetic resonance images (MRI) is a difficult task due to various tumour types. Using information and features from multimodal MRI including structural MRI and isotropic (p) and anisotropic (q) components derived from the diffusion tensor imaging (DTI) may result in a more accurate analysis of brain images. METHODS We propose a novel 3D supervoxel based learning method for segmentation of tumour in multimodal MRI brain images (conventional MRI and DTI). Supervoxels are generated using the information across the multimodal MRI dataset. For each supervoxel, a variety of features including histograms of texton descriptor, calculated using a set of Gabor filters with different sizes and orientations, and first order intensity statistical features are extracted. Those features are fed into a random forests (RF) classifier to classify each supervoxel into tumour core, oedema or healthy brain tissue. RESULTS The method is evaluated on two datasets: 1) Our clinical dataset: 11 multimodal images of patients and 2) BRATS 2013 clinical dataset: 30 multimodal images. For our clinical dataset, the average detection sensitivity of tumour (including tumour core and oedema) using multimodal MRI is 86% with balanced error rate (BER) 7%; while the Dice score for automatic tumour segmentation against ground truth is 0.84. The corresponding results of the BRATS 2013 dataset are 96%, 2% and 0.89, respectively. CONCLUSION The method demonstrates promising results in the segmentation of brain tumour. Adding features from multimodal MRI images can largely increase the segmentation accuracy. The method provides a close match to expert delineation across all tumour grades, leading to a faster and more reproducible method of brain tumour detection and delineation to aid patient management.