Inter-rater agreement in glioma segmentations on longitudinal MRI.

BACKGROUND: Tumor segmentation of glioma on MRI is a technique to monitor, quantify and report disease progression. Manual MRI segmentation is the gold standard but very labor intensive. At present the quality of this gold standard is not known for different stages of the disease, and prior work has mainly focused on treatment-naive glioblastoma. In this paper we studied the inter-rater agreement of manual MRI segmentation of glioblastoma and WHO grade II-III glioma for novices and experts at three stages of disease. We also studied the impact of inter-observer variation on extent of resection and growth rate. METHODS: In 20 patients with WHO grade IV glioblastoma and 20 patients with WHO grade II-III glioma (defined as non-glioblastoma) both the enhancing and non-enhancing tumor elements were segmented on MRI, using specialized software, by four novices and four experts before surgery, after surgery and at time of tumor progression. We used the generalized conformity index (GCI) and the intra-class correlation coefficient (ICC) of tumor volume as main outcome measures for inter-rater agreement. RESULTS: For glioblastoma, segmentations by experts and novices were comparable. The inter-rater agreement of enhancing tumor elements was excellent before surgery (GCI 0.79, ICC 0.99) poor after surgery (GCI 0.32, ICC 0.92), and good at progression (GCI 0.65, ICC 0.91). For non-glioblastoma, the inter-rater agreement was generally higher between experts than between novices. The inter-rater agreement was excellent between experts before surgery (GCI 0.77, ICC 0.92), was reasonable after surgery (GCI 0.48, ICC 0.84), and good at progression (GCI 0.60, ICC 0.80). The inter-rater agreement was good between novices before surgery (GCI 0.66, ICC 0.73), was poor after surgery (GCI 0.33, ICC 0.55), and poor at progression (GCI 0.36, ICC 0.73). Further analysis showed that the lower inter-rater agreement of segmentation on postoperative MRI could only partly be explained by the smaller volumes and fragmentation of residual tumor. The median interquartile range of extent of resection between raters was 8.3% and of growth rate was 0.22 mm/year. CONCLUSION: Manual tumor segmentations on MRI have reasonable agreement for use in spatial and volumetric analysis. Agreement in spatial overlap is of concern with segmentation after surgery for glioblastoma and with segmentation of non-glioblastoma by non-experts.

Optimizing the association between disability and biological markers in MS.

OBJECTIVE: Axonal damage is an important feature of MS pathology and the likely substrate of development of progressive disability. Brain volume measurement on MRI can be used as an overall marker of tissue damage and axonal loss. The authors studied the relation of brain volume measurements with the MS Functional Composite (MSFC) in an attempt to improve the clinico-radiologic association. METHODS: In 137 patients with MS (80 relapsing-remitting [RR], 36 secondary progressive [SP], and 21 primary progressive [PP]) and 12 healthy controls, a brain MRI scan was obtained. Patients also underwent MSFC and Expanded Disability Status Scale (EDSS) assessments. MRI analysis included determination of hypointense T1- and hyperintense T2-weighted lesion load, and two brain volume measurements: 1) the parenchymal fraction (PF): whole brain parenchyma/intracranial volume; and 2) the ventricular fraction (VF): ventricular volume/whole brain parenchyma. RESULTS: The median PF was smaller and the median VF larger in the patient group (0.81 for PF and 0.029 for VF) than in the control group (0.87 for PF, p < 0.001; and 0.013 for VF, p < 0.01). For the patient population, moderate correlations were found between brain volume measurements and MSFC (0.36 for PF and -0.40 for VF). Patients with short disease duration showed a correlation of MSFC with both brain and lesion volume measurements on MRI, whereas patients with long disease duration only showed a correlation with brain volume measurements. CONCLUSION: Brain volume measurements are correlated with disability as assessed by the MSFC. Although in the early phase of the disease the amount of focal demyelination is important, the residual brain volume seems to be more relevant in determining disability in later phases of the disease.

Axonal damage in the spinal cord of MS patients occurs largely independent of T2 MRI lesions.

OBJECTIVE: To determine the degree of axonal damage in relationship to signal abnormalities on T2-weighted high-resolution MRI in spinal cord tissue of patients with MS. METHODS: Spinal cord specimens of nine patients with MS and four controls were imaged at high resolution (4.7 T) in an axial plane and scored for lesions with increased signal intensity (SI). Histopathologic sections were cut and immunostained with NE14 (neurofilament marker) and Luxol fast blue (myelin stain). For each area, axonal density and diameter were quantified; axonal irregularity, NE14 axonal staining intensity, and myelin content were semiquantitatively scored. Included were 209 areas from MS cases and 109 areas from control cases distributed over lateral, posterior, and anterior columns. RESULTS: In control cases, no SI changes were found, average density of axons was 26,989/mm(2), average diameter was 1.1 micro m, and all scores for axonal irregularity, NE14 staining intensity, and myelin were normal. In MS cases, areas with increased SI were found, average axonal density was 11,807/mm(2) (p < 0.0001), and average axonal diameter 2.0 micro m (p = 0.001). Areas with high SI on MRI had lowest axonal density (average count: 10,504/mm(2); range: 3,433 to 26,325/mm(2)), largest diameter (average: 2.3 micro m; range: 1.0 to 4.0 micro m), and highest axonal irregularity and NE14 staining intensity compared to normal appearing cord tissue (NACT). However, NACT of MS cases also had lower axonal density (14,158/mm(2)) and higher average axonal diameter (1.6 micro m) than controls. CONCLUSIONS: Marked axonal loss occurs in MS spinal cords, largely independent of the degree of signal abnormality on T2-weighted MRI.

The predictive value of 3D time-of-flight MR angiography in assessment of brain arteriovenous malformation obliteration after radiosurgery.

BACKGROUND AND PURPOSE: The purpose of radiosurgery of bAVMs is complete angiographic obliteration of its nidus. We assessed the diagnostic accuracy of 1.5T T2-weighted MR imaging and TOF-MRA images for detecting nidus obliteration after radiosurgery. MATERIALS AND METHODS: The pre- and postradiosurgery MR images and DSA images from 120 patients who were radiosurgically treated for a bAVM were re-evaluated by 2 observers for patency of the nidus (preradiosurgery) and obliteration (postradiosurgery: final follow-up MR imaging), by using a 3-point scale of confidence. Consensus reading of the DSA after radiosurgery was considered the criterion standard for obliteration. Sensitivity, specificity, PPVs, and NPVs, and overall diagnostic performance by using ROC were determined. RESULTS: Mean bAVM volume during radiosurgery was 3.4 mL (95% CI, 2.6-4.3 mL). Sixty-six patients (55%) had undergone previous endovascular embolization. The mean intervals between radiosurgery and follow-up MR imaging and for DSA, respectively, were 35.6 months (95% CI, 32.3-38.9 months) and 42.1 months (95% CI, 40.3-44.0 months). With ROC, an area under curve of 0.81-0.83 was found. PPVs of final follow-up MR-imaging for definitive obliteration varied between 0.89 [corrected] and 0.95. NPV was 0.52 [corrected] . An average false-positive rate, meaning overestimation of nidus obliteration of 0.10 [corrected] and an average false-negative rate, meaning underestimation of nidus obliteration of 0.42 [corrected] were found. CONCLUSIONS: MRA is insufficient to diagnose obliteration in the follow-up of bAVMs after radiosurgery. A remaining nidus diameter <10 mm seems to be the major limiting factor for reliable assessment of obliteration. We highly recommend follow-up DSA for definitive diagnosis of complete obliteration.

Spinal cord grey matter lesions in multiple sclerosis detected by post-mortem high field MR imaging.

BACKGROUND: Post-mortem studies demonstrate extensive grey matter demyelination in MS, both in the brain and in the spinal cord. However the clinical significance of these plaques is unclear, largely because they are grossly underestimated by MR imaging at conventional field strengths. Indeed post-mortem MR studies suggest the great majority of lesions in the cerebral cortex go undetected, even when performed at high field. Similar studies have not been performed using post-mortem spinal cord material. AIM: To assess the sensitivity of high field post-mortem MRI for detecting grey matter lesions in the spinal cord in MS. METHODS: Autopsy material was obtained from 11 MS cases and 2 controls. Proton Density-weighted images of this formalin-fixed material were acquired at 4.7 Tesla before the tissue was sectioned and stained for Myelin Basic Protein. Both the tissue sections and the MR images were scored for grey matter and white matter plaques, with the readers of the MR images being blinded to the histopathology results. RESULTS: Our results indicate that post-mortem imaging at 4.7 Tesla is highly sensitive for cord lesions, detecting 87% of white matter lesions and 73% of grey matter lesions. The MR changes were highly specific for demyelination, with all lesions scored on MRI corresponding to areas of demyelination. CONCLUSION: Our work suggests that spinal cord grey matter lesions may be detected on MRI more readily than GM lesions in the brain, making the cord a promising site to study the functional consequences of grey matter demyelination in MS.

At the heart of primary progressive multiple sclerosis: three cases with diffuse MRI abnormalities only.

We present three patients with a clinical course and cerebrospinal fluid findings consistent with a diagnosis of primary progressive multiple sclerosis (PPMS). Extensive and repeated magnetic resonance imaging (MRI) examinations showed only diffuse abnormality in brain and spinal cord, but no focal lesions. We propose that these cases represent the most pure form of PPMS, even though according to currently applied criteria this diagnosis can not be made in the absence of focal lesions on MRI.

Spinal cord abnormalities in recently diagnosed MS patients: added value of spinal MRI examination.

OBJECTIVE: The most recent diagnostic criteria for multiple sclerosis (MS) ascertain that findings from spinal cord MRI can be used to demonstrate dissemination in space. Because little is known about the prevalence and characteristics of cord lesions early in the disease, the authors studied the prevalence of spinal cord abnormalities in patients with early-stage MS and assessed their impact on diagnostic classification. METHODS: The brains and spinal cords of 104 recently diagnosed patients with MS were examined. Median interval between first symptom and diagnosis was 18.4 months. The brain MRI protocol included before and after gadolinium axial T1-weighted conventional spin-echo sequences and dual-echo spin-echo images. For spinal cord MRI, sagittal cardiac-triggered dual-echo T2-weighted and sagittal T1-weighted spin-echo images were included. Clinical assessment for each patient included age, sex, clinical signs for spinal cord involvement, and Expanded Disability Status Scale. RESULTS: Abnormal cord MRIs were found in 83% of patients, usually with only focal lesions. Diffuse cord abnormalities were found in 13% of patients, although in isolation they were found in only three patients. Focal cord lesions were often multiple (median number, 3.0), small (median, 0.8 vertebral segments), and primarily (56.4%) situated in the cervical spinal cord. In 68 of 104 patients (65.4%), two or more focal lesions were visible on spinal cord images. The criteria for dissemination in space, as defined in the McDonald criteria for the brain, were met in only 66.3% of the patients. This percentage increased to 84.6% when spinal cord MRI abnormalities were also included. CONCLUSION: Spinal cord abnormalities are prevalent in patients with early-stage MS, have distinct morphologic characteristics, and help to determine dissemination in space at time of diagnosis.

Comparison of a conventional cardiac-triggered dual spin-echo and a fast STIR sequence in detection of spinal cord lesions in multiple sclerosis.

The current optimal imaging protocol in spinal cord MR imaging in patients with multiple sclerosis includes a long TR conventional spin-echo (CSE) sequence, requiring long acquisition times. Using short tau inversion recovery fast spin-echo (fast STIR) sequences both acquisition time can be shortened and sensitivity in the detection of multiple sclerosis (MS) abnormalities can be increased. This study compares both sequences for the potential to detect both focal and diffuse spinal abnormalities. Spinal cords of 5 volunteers and 20 MS patients were studied at 1.0 T. Magnetic resonance imaging included cardiac-gated sagittal dual-echo CSE and a cardiac-gated fast STIR sequence. Images were scored regarding number, size, and location of focal lesions, diffuse abnormalities and presence/hindrance of artifacts by two experienced radiologists. Examinations were scored as being definitely normal, indeterminate, or definitely abnormal. Interobserver agreement regarding focal lesions was higher for CSE (kappa = 0.67) than for fast STIR (kappa = 0.57) but did not differ significantly. Of all focal lesions scored in consensus, 47% were scored on both sequences, 31% were only detected by fast STIR, and 22% only by dual-echo CSE (n.s.). Interobserver agreement for diffuse abnormalities was lower with fast STIR (kappa = 0.48) than dual-echo CSE (kappa = 0.65; n. s.). After consensus, fast STIR showed in 10 patients diffuse abnormalities and dual-echo CSE in 3. After consensus, in 19 of 20 patients dual-echo CSE scans were considered as definitely abnormal compared with 17 for fast STIR. The fast STIR sequence is a useful adjunct to dual-echo CSE in detecting focal abnormalities and is helpful in detecting diffuse MS abnormalities in the spinal cord. Due to the frequent occurrence of artifacts and the lower observer concordance, fast STIR cannot be used alone.