No abstract

Abstract-Link prediction is a fundamental problem in social network analysis and modern-day commercial applications such as Facebook and Myspace. Most existing research approaches this problem by exploring the topological structure of a social network using only one source of information. However, in many application domains, in addition to the social network of interest, there are a number of auxiliary social networks and/or derived proximity networks available. The contribution of the paper is twofold: (1) a supervised learning framework that can effectively and efficiently learn the dynamics of social networks in the presence of auxiliary networks; (2) a feature design scheme for constructing a rich variety of path-based features using multiple sources, and an effective feature selection strategy based on structured sparsity. Extensive experiments on three real-world collaboration networks show that our model can effectively learn to predict new links using multiple sources, yielding higher prediction accuracy than unsupervised and singlesource supervised models.

No abstract

No abstract

ObjectivesThe aim of this study is to comprehensively evaluate the advantage of diffusion kurtosis imaging (DKI) in distinguishing pathological complete response (pCR) from non-pCR patients with locally advanced rectal cancer (LARC) after neoadjuvant chemoradiation therapy (CRT) in comparison to conventional diffusion-weighted imaging (DWI).Material and MethodsFifty-six consecutive patients diagnosed with LARC were prospectively enrolled and underwent pre- and post-CRT MRI on a 3.0 T MRI scanner. Apparent diffusion coefficient (ADC), mean diffusion (MD) and mean kurtosis (MK) values of the tumor were measured in pre- and post-CRT phases and then compared to histopathologic findings after total mesorectal excision (TME). Both Mann-Whitney U-test and Kruskal-Wallis test were used as statistical methods. Diagnostic performance was determined using receiver operating characteristic (ROC) curve analysis.ResultsFor a total of 56 rectal lesions (pCR, n = 14; non-pCR, n = 42), the MKpre and MKpost values were much lower for the pCR patients (mean±SD, 0.72±0.09 and 0.56±0.06, respectively) than those for the non-pCR patients (0.89±0.11 and 0.68±0.08, respectively) (p < 0.001). The ADCpost and the change ratio of apparent diffusion coefficient (ADCratio) values was significantly higher for the pCR patients (mean±SD, 1.31±0.13 and 0.64±0.34, respectively) than for the non-pCR patients (1.12±0.16 and 0.33±0.27, respectively) (p < 0.001 and p = 0.001, respectively). In addition, the MDpost and the change ratio of mean diffusion (MDratio) (2.45±0.33 vs. 1.95±0.30, p < 0.001; 0.80±0.43 vs. 0.35±0.32, p < 0.001, respectively) also increased, whereas the ADCpre, MDpre and the change ratio of mean kurtosis (MKratio) of the pCR (0.82±0.11, 1.40±0.21, and 0.23±0.010, respectively) exhibited a neglectable difference with that of the non-pCR (p = 0.332, 0.269, and 0.678, respectively). The MKpost showed relatively high sensitivity (92.9%) and high specificity (83.3%) in comparison to other image indices. The area under the receiver operating characteristic curve (AUROC) that is available for the assessment of pCR using MKpost (0.908, cutoff value = 0.6196) were larger than other parameters and the overall accuracy of MKpost (85.7%) was the highest.ConclusionsBoth DKI and conventional DWI hold great potential in predicting treatment response to neoadjuvant chemoradiation therapy in rectal cancer. The DKI parameters, especially MKpost, showed a higher specificity than conventional DWI in assessing pCR and non-pCR in patients with LARC, but the pre-CRT ADC and MD are unreliable.