Raj Jena

The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)

In this paper we report the set-up and results of the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) organized in conjunction with the MICCAI 2012 and 2013 conferences. Twenty state-of-the-art tumor segmentation algorithms were applied to a set of 65 multi-contrast MR scans of low- and high-grade glioma patients - manually annotated by up to four raters - and to 65 comparable scans generated using tumor image simulation software. Quantitative evaluations revealed considerable disagreement between the human raters in segmenting various tumor sub-regions (Dice scores in the range 74%-85%), illustrating the difficulty of this task. We found that different algorithms worked best for different sub-regions (reaching performance comparable to human inter-rater variability), but that no single algorithm ranked in the top for all sub-regions simultaneously. Fusing several good algorithms using a hierarchical majority vote yielded segmentations that consistently ranked above all individual algorithms, indicating remaining opportunities for further methodological improvements. The BRATS image data and manual annotations continue to be publicly available through an online evaluation system as an ongoing benchmarking resource.

Decision forests for tissue-specific segmentation of high-grade gliomas in multi-channel MR.

We present a method for automatic segmentation of high-grade gliomas and their subregions from multi-channel MR images. Besides segmenting the gross tumor, we also differentiate between active cells, necrotic core, and edema. Our discriminative approach is based on decision forests using context-aware spatial features, and integrates a generative model of tissue appearance, by using the probabilities obtained by tissue-specific Gaussian mixture models as additional input for the forest. Our method classifies the individual tissue types simultaneously, which has the potential to simplify the classification task. The approach is computationally efficient and of low model complexity. The validation is performed on a labeled database of 40 multi-channel MR images, including DTI. We assess the effects of using DTI, and varying the amount of training data. Our segmentation results are highly accurate, and compare favorably to the state of the art.

A mathematical model of brain tumour response to radiotherapy and chemotherapy considering radiobiological aspects

Glioblastoma is the most frequent and malignant brain tumour. For many years, the conventional treatment has been maximal surgical resection followed by radiotherapy (RT), with a median survival time of less than 10 months. Previously, the use of adjuvant chemotherapy (given after RT) has failed to demonstrate a statistically significant survival advantage. Recently, a randomized phase III trial has confirmed the benefit of temozolomide (TMZ) and has defined a new standard of care for the treatment of patients with high-grade brain tumours. The results showed an increase of 2.5 months in median survival, and 16.1% in 2 year survival, for patients receiving RT with TMZ compared with RT alone. It is not clear whether the major benefit of TMZ comes from either concomitant administration of TMZ with RT, or from six cycles of adjuvant TMZ, or both. The objectives were to develop our original model, which addressed survival after RT, to construct a new module to assess the potential role of TMZ from clinical data, and to explore its synergistic contribution in addition to radiation. The model has been extended to include radiobiological parameters. The addition of the linear quadratic equation to describe cellular response to treatment has enabled us to quantify the effects of radiation and TMZ in radiobiological terms. The results indicate that the model achieves an excellent fit to the clinical data, with the assumption that TMZ given concomitantly with RT synergistically increases radiosensitivity. The alternative, that the effect of TMZ is due only to direct cell killing, does not fit the clinical data so well. The addition of concomitant TMZ appears to change the radiobiological parameters. This aspect of our results suggests possible treatment developments. Our observations need further evaluations in real clinical trials, may suggest treatment strategies for new trials, and inform their design.

Relationship between irradiated breast volume and late normal tissue complications: A systematic review

The concept of radiation dose-volume effect has been exploited in breast cancer as boost treatment for high risk patients and more recently in trials of Partial Breast Irradiation for low risk patients. However, there appears to be paucity of published data on the dose-volume effect of irradiation on breast tissue including the recently published report on Quantitative Analyses of Normal Tissue Effects in the Clinic (QUANTEC). This systematic review looks at the current literature for relationship between irradiated breast volume and normal tissue complications and introduces the concept of dose modulation.

Optimal dosing of cancer chemotherapy using model predictive control and moving horizon state/parameter estimation

Model predictive control (MPC), originally developed in the community of industrial process control, is a potentially effective approach to optimal scheduling of cancer therapy. The basis of MPC is usually a state-space model (a system of ordinary differential equations), whereby existing studies usually assume that the entire states can be directly measured. This paper aims to demonstrate that when the system states are not fully measurable, in conjunction with model parameter discrepancy, MPC is still a useful method for cancer treatment. This aim is achieved through the application of moving horizon estimation (MHE), an optimisation-based method to jointly estimate the system states and parameters. The effectiveness of the MPC-MHE scheme is illustrated through scheduling the dose of tamoxifen for simulated tumour-bearing patients, and the impact of estimation horizon and magnitude of parameter discrepancy is also investigated.

In vitro evaluation of combined temozolomide and radiotherapy using X-rays and high-linear energy transfer radiation for glioblastoma

High-linear energy transfer radiation offers superior biophysical properties over conventional radiotherapy and may have a great potential for treating radioresistant tumors, such as glioblastoma. However, very little pre-clinical data exists on the effects of high-LET radiation on glioblastoma cell lines and on the concomitant application of chemotherapy. This study investigates the in vitro effects of temozolomide in combination with low-energy protons and α particles. Cell survival, DNA damage and repair, and cell growth were examined in four human glioblastoma cell lines (LN18, T98G, U87 and U373) after treatment with either X rays, protons (LET 12.91 keV/μm), or α particles (LET 99.26 keV/μm) with or without concurrent temozolomide at clinically-relevant doses of 25 and 50 μM. The relative biological effectiveness at 10% survival (RBE10) increased as LET increased: 1.17 and 1.06 for protons, and 1.84 and 1.68 for α particles in the LN18 and U87 cell lines, respectively. Temozolomide administration increased cell killing in the O6-methylguanine DNA methyltransferase-methylated U87 and U373 cell lines. In contrast, temozolomide provided no therapeutic enhancement in the methylguanine DNA methyltransferase-unmethylated LN18 and T98G cell lines. In addition, the residual number of γ-H2AX foci at 24 h after treatment with radiation and concomitant temozolomide was found to be lower than or equal to that expected by DNA damage with either of the individual treatments. Kinetics of foci disappearance after X-ray and proton irradiation followed similar time courses; whereas, loss of γ-H2AX foci after α particle irradiation occurred at a slower rate than that by low-LET radiation (half-life 12.51–16.87 h). The combination of temozolomide with different radiation types causes additive rather than synergistic cytotoxicity. Nevertheless, particle therapy combined with chemotherapy may offer a promising alternative with the additional benefit of superior biophysical properties. It is also possible that new fractionation schedules could be designed to exploit the change in DNA repair kinetics when MGMT-methylated cells respond to high-LET radiation.

Accumulated dose to the rectum, measured using dose–volume histograms and dose-surface maps, is different from planned dose in all patients treated with radiotherapy for prostate cancer

Objective: We sought to calculate accumulated dose (DA) to the rectum in patients treated with radiotherapy for prostate cancer. We were particularly interested in whether dose–surface maps (DSMs) provide additional information to dose–volume histograms (DVHs). Methods: Manual rectal contours were obtained for kilovoltage and daily megavoltage CT scans for 10 participants from the VoxTox study (380 scans). Daily delivered dose recalculation was performed using a ray-tracing algorithm. Delivered DVHs were summated to create accumulated DVHs. The rectum was considered as a cylinder, cut and unfolded to produce daily delivered DSMs; these were summated to produce accumulated DSMs. Results: Accumulated dose-volumes were different from planned in all participants. For one participant, all DA levels were higher and all volumes were larger than planned. For four participants, all DA levels were lower and all volumes were smaller than planned. For each of these four participants, ≥1% of pixels on the accumulated DSM received ≥5 Gy more than had been planned. Conclusion: Differences between accumulated and planned dose-volumes were seen in all participants. DSMs were able to identify differences between DA and planned dose that could not be appreciated from the DVHs. Further work is needed to extract the dose data embedded in the DSMs. These will be correlated with toxicity as part of the VoxTox Programme. Advances in knowledge: DSMs are able to identify differences between DA and planned dose that cannot be appreciated from DVHs alone and should be incorporated into future studies investigating links between DA and toxicity.

Bevacizumab in Neurofibromatosis type 2 (NF2) related vestibular schwannomas: a nationally coordinated approach to delivery and prospective evaluation

Background NF2 patients develop multiple nervous system tumors including bilateral vestibular schwannomas (VS). The tumors and their surgical treatment are associated with deafness, neurological disability, and mortality.Medical treatment with bevacizumab has been reported to reduce VS growth and to improve hearing. In addition to evaluating these effects, this study also aimed to determine other important consequences of treatment including patient-reported quality of life and the impact of treatment on surgical VS rates. Methods Patients treated with bevacizumab underwent serial prospective MRI, audiology, clinical, CTCAE-4.0 adverse events, and NFTI-QOL quality-of-life assessments. Tumor volumetrics were classified according to the REiNs criteria and annual VS surgical rates reviewed. Results Sixty-one patients (59% male), median age 25 years (range, 10-57), were reviewed. Median follow-up was 23 months (range, 3-53). Partial volumetric tumor response (all tumors) was seen in 39% and 51% had stabilization of previously growing tumors. Age and pretreatment growth rate were predictors of response. Hearing was maintained or improved in 86% of assessable patients. Mean NFTI-QOL scores improved from 12.0 to 10.7 (P < .05). Hypertension was observed in 30% and proteinuria in 16%. Twelve treatment breaks occurred due to adverse events. The rates of VS surgery decreased after the introduction of bevacizumab. Conclusion Treatment with bevacizumab in this large, UK-wide cohort decreased VS growth rates and improved hearing and quality of life. The potential risk of surgical iatrogenic damage was also reduced due to an associated reduction in VS surgical rates. Ongoing follow-up of this cohort will determine the long-term benefits and risks of bevacizumab treatment.

Classification Forests for Semantic Segmentation of Brain Lesions in Multi-channel MRI

Classification forests, as discussed in Chapter 2, have a series of advantageous properties which make them a very good choice for applications in medical image analysis. Classification forests are inherent multi-label classifiers (which allows for the simultaneous segmentation of different tissues), have good generalization properties (which is important as training data is often scarce in medical applications), and are able to deal with very high-dimensional feature spaces (which allows the use of non-local and context-aware features to describe the input data). In this chapter we demonstrate how classification forests can be used as a basic building block to develop state of the art systems for medical image analysis in two challenging applications. These applications perform the segmentation of two different types of brain lesions based on 3D multi-channel magnetic resonance images (MRI) as input. More specifically, we discuss (1) the segmentation of the individual tissues of high-grade brain tumor lesions, and (2) the segmentation of multiple-sclerosis lesions.

A multicentre study of the evidence for customized margins in photon breast boost radiotherapy

Objective: To determine if subsets of patients may benefit from smaller or larger margins when using laser setup and bony anatomy verification of breast tumour bed (TB) boost radiotherapy (RT). Methods: Verification imaging data acquired using cone-beam CT, megavoltage CT or two-dimensional kilovoltage imaging on 218 patients were used (1574 images). TB setup errors for laser-only setup (dlaser) and for bony anatomy verification (dbone) were determined using clips implanted into the TB as a gold standard for the TB position. Cases were grouped by centre-, patient- and treatment-related factors, including breast volume, TB position, seroma visibility and surgical technique. Systematic (Σ) and random (σ) TB setup errors were compared between groups, and TB planning target volume margins (MTB) were calculated. Results: For the study population, Σlaser was between 2.8 and 3.4 mm, and Σbone was between 2.2 and 2.6 mm, respectively. Females with larger breasts (p = 0.03), easily visible seroma (p ≤ 0.02) and open surgical technique (p ≤ 0.04) had larger Σlaser. Σbone was larger for females with larger breasts (p = 0.02) and lateral tumours (p = 0.04). Females with medial tumours (p < 0.01) had smaller Σbone. Conclusion: If clips are not used, margins should be 8 and 10 mm for bony anatomy verification and laser setup, respectively. Individualization of TB margins may be considered based on breast volume, TB and seroma visibility. Advances in knowledge: Setup accuracy using lasers and bony anatomy is influenced by patient and treatment factors. Some patients may benefit from clip-based image guidance more than others.