Evaluation of current algorithms for segmentation of scar tissue from late gadolinium enhancement cardiovascular magnetic resonance of the left atrium: an open-access grand challenge.

BACKGROUND: Late Gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging can be used to visualise regions of fibrosis and scarring in the left atrium (LA) myocardium. This can be important for treatment stratification of patients with atrial fibrillation (AF) and for assessment of treatment after radio frequency catheter ablation (RFCA). In this paper we present a standardised evaluation benchmarking framework for algorithms segmenting fibrosis and scar from LGE CMR images. The algorithms reported are the response to an open challenge that was put to the medical imaging community through an ISBI (IEEE International Symposium on Biomedical Imaging) workshop. METHODS: The image database consisted of 60 multicenter, multivendor LGE CMR image datasets from patients with AF, with 30 images taken before and 30 after RFCA for the treatment of AF. A reference standard for scar and fibrosis was established by merging manual segmentations from three observers. Furthermore, scar was also quantified using 2, 3 and 4 standard deviations (SD) and full-width-at-half-maximum (FWHM) methods. Seven institutions responded to the challenge: Imperial College (IC), Mevis Fraunhofer (MV), Sunnybrook Health Sciences (SY), Harvard/Boston University (HB), Yale School of Medicine (YL), King's College London (KCL) and Utah CARMA (UTA, UTB). There were 8 different algorithms evaluated in this study. RESULTS: Some algorithms were able to perform significantly better than SD and FWHM methods in both pre- and post-ablation imaging. Segmentation in pre-ablation images was challenging and good correlation with the reference standard was found in post-ablation images. Overlap scores (out of 100) with the reference standard were as follows: Pre: IC = 37, MV = 22, SY = 17, YL = 48, KCL = 30, UTA = 42, UTB = 45; Post: IC = 76, MV = 85, SY = 73, HB = 76, YL = 84, KCL = 78, UTA = 78, UTB = 72. CONCLUSIONS: The study concludes that currently no algorithm is deemed clearly better than others. There is scope for further algorithmic developments in LA fibrosis and scar quantification from LGE CMR images. Benchmarking of future scar segmentation algorithms is thus important. The proposed benchmarking framework is made available as open-source and new participants can evaluate their algorithms via a web-based interface.

The precision and sensitivity of (18)F-fluoride PET for measuring regional bone metabolism: a comparison of quantification methods.

UNLABELLED: The planning of research studies requires an understanding of the minimum number of subjects required. The aim of this study was to evaluate different methods of analyzing (18)F-fluoride PET ((18)F(-) PET) dynamic spine scans to find the approach that requires the smallest sample size to detect a statistically significant response to treatment. METHODS: Eight different approaches to (18)F(-) PET analysis (3 variants of the Hawkins 3-tissue compartmental model, 3 variants of spectral analysis, deconvolution, and Patlak analysis) were used to evaluate the fluoride plasma clearance to bone mineral (K(i)). Standardized uptake values (SUVs) were also studied. Data for 20 women who had (18)F(-) PET spine scans at 0, 6, and 12 mo after stopping long-term bisphosphonate treatment were used to compare precision errors. Data for 18 women who had scans at baseline and 6 mo after starting teriparatide treatment were used to compare response to treatment. RESULTS: The 4 approaches that fitted the rate constant k(4) describing the reverse flow of (18)F from bone as a free variable showed close agreement in K(i) values, with correlation coefficients greater than 0.97. Their %CVs were 14.4%-14.8%, and treatment response to teriparatide was 23.2%-23.8%. The 3 methods that assumed k(4) = 0 gave K(i) values 20%-25% lower than the other methods, with correlation coefficients of 0.83-0.94, percentage coefficients of variation (%CVs) of 12.9%-13.3%, and treatment response of 25.2%-28.3%. A Hawkins model with k(4) = 0.01 min(-1) did not perform any better (%CV, 14.2%; treatment response, 26.1%). Correlation coefficients between SUV and the different K(i) methods varied between 0.60 and 0.65. Although SUV gave the best precision (%CV, 10.1%), the treatment response (3.1%) was not statistically significant. CONCLUSION: Methods that calculated K(i) assuming k(4) = 0 required fewer subjects to demonstrate a statistically significant response to treatment than methods that fitted k(4) as a free variable. Although SUV gave the smallest precision error, the absence of any significant changes make it unsuitable for examining response to treatment in this study.