To ensure undisrupted business, large Internet companies need to closely monitor various KPIs (e.g., Page Views, number of online users, and number of orders) of its Web applications, to accurately detect anomalies and trigger timely troubleshooting/mitigation. However, anomaly detection for these seasonal KPIs with various patterns and data quality has been a great challenge, especially without labels. In this paper, we proposed Donut, an unsupervised anomaly detection algorithm based on VAE. Thanks to a few of our key techniques, Donut greatly outperforms a state-of-arts supervised ensemble approach and a baseline VAE approach, and its best F-scores range from 0.75 to 0.9 for the studied KPIs from a top global Internet company. We come up with a novel KDE interpretation of reconstruction for Donut, making it the first VAE-based anomaly detection algorithm with solid theoretical explanation.

Purpose To develop a three-dimensional (3D) high-spatial-resolution time-efficient sequence for use in quantitative vessel wall T1 mapping. Materials and Methods A previously described sequence, simultaneous noncontrast angiography and intraplaque hemorrhage (SNAP) imaging, was extended by introducing 3D golden angle radial k-space sampling (GOAL-SNAP). Sliding window reconstruction was adopted to reconstruct images at different inversion delay times (different T1 contrasts) for voxelwise T1 fitting. Phantom studies were performed to test the accuracy of T1 mapping with GOAL-SNAP against a two-dimensional inversion recovery (IR) spin-echo (SE) sequence. In vivo studies were performed in six healthy volunteers (mean age, 27.8 years ± 3.0 [standard deviation]; age range, 24–32 years; five male) and five patients with atherosclerosis (mean age, 66.4 years ± 5.5; range, 60–73 years; five male) to compare T1 measurements between vessel wall sections (five per artery) with and without intraplaque hemorrhage (IPH). Statistical analyses included Pearson correlation coefficient, Bland-Altman analysis, and Wilcoxon rank-sum test with data permutation by subject. Results Phantom T1 measurements with GOAL-SNAP and IR SE sequences showed excellent correlation (R2 = 0.99), with a mean bias of −25.8 msec ± 43.6 and a mean percentage error of 4.3% ± 2.5. Minimum T1 was significantly different between sections with IPH and those without it (mean, 371 msec ± 93 vs 944 msec ± 120; P = .01). Estimated T1 of normal vessel wall and muscle were 1195 msec ± 136 and 1117 msec ± 153, respectively. Conclusion High-spatial-resolution (0.8 mm isotropic) time-efficient (5 minutes) vessel wall T1 mapping is achieved by using the GOAL-SNAP sequence. This sequence may yield more quantitative reproducible biomarkers with which to characterize IPH and monitor its progression.

The purpose of this study was to detect the physiological process of FDG's filtration from blood to urine and to establish a mathematical model to describe the process. Dynamic positron emission tomography scan for FDG was performed on seven normal volunteers. The filtration process in kidney can be seen in the sequential images of each study. Variational distribution of FDG in kidney can be detected in dynamic data. According to the structure and function, kidney is divided into parenchyma and pelvis. A unidirectional three-compartment model is proposed to describe the renal function in FDG excretion. The time-activity curves that were picked up from the parenchyma, pelvis, and abdominal aorta were used to estimate the parameter of the model. The output of the model has fitted well with the original curve from dynamic data.

The annual progression of carotid wall volume is independently associated with recurrent ischemic cerebrovascular events, and this measurement has added value for intraplaque hemorrhage and fibrous cap rupture in predicting future events.