Cascaded Algorithm Selection With Extreme-Region UCB Bandit.

AutoML aims at best configuring learning systems automatically. It contains core subtasks of algorithm selection and hyper-parameter tuning. Previous approaches considered searching in the joint hyper-parameter space of all algorithms, which forms a huge but redundant space and causes an inefficient search. We tackle this issue in a cascaded algorithm selection way, which contains an upper-level process of algorithm selection and a lower-level process of hyper-parameter tuning for algorithms. While the lower-level process employs an anytime tuning approach, the upper-level process is naturally formulated as a multi-armed bandit, deciding which algorithm should be allocated one more piece of time for the lower-level tuning. To achieve the goal of finding the best configuration, we propose the Extreme-Region Upper Confidence Bound (ER-UCB) strategy. Unlike UCB bandits that maximize the mean of feedback distribution, ER-UCB maximizes the extreme-region of feedback distribution. We first consider stationary distributions and propose the ER-UCB-S algorithm that has O(Klnn) regret upper bound with K arms and n trials. We then extend to non-stationary settings and propose the ER-UCB-N algorithm that has O(Knν) regret upper bound, where [Formula: see text]. Finally, empirical studies on synthetic and AutoML tasks verify the effectiveness of ER-UCB-S/N by their outperformance in corresponding settings.

Evaluation of suspicious breast lesions with diffusion kurtosis MR imaging and connection with prognostic factors.

PURPOSE: To investigate the additional value of DKI in discriminating suspicious breast lesions on DCE-MRI, as compared with conventional DWI; and to explore connection between DKI-parameters and prognostic factors of breast cancers. METHODS: The institutional review board approved this retrospective study and written informed consent was waived. Totally, 300 women (mean age, 43.2 ± 10.4 years) with suspicious breast lesions on DCE-MRI were enrolled from November 2014 to September 2019. With pathology as reference, performance of ADC, Kapp and Dapp in discriminating suspicious breast lesions were analyzed by receiver operating characteristic (ROC) analysis with area under ROC curve (AUC). The specificities of parameters were compared by Chi-square test. The ADC, Kapp and Dapp of breast cancers with different receptor status were compared using Student's t or Mann-Whitney U or Kruskal-Wallis test. RESULTS: There were 344 suspicious breast lesions (220 malignant, 124 benign) in 300 women. No significant differences were found for AUCs of ADC and DKI-parameters in discriminating suspicious breast lesions (0.882 vs. 0.888, p = 0.480). The specificities were significantly higher with ADC and Dapp than that with DCE-MRI (p = 0.003 and 0.005). The ADC, Kapp and Dapp were correlated with HER2 expression and lymph node status, and ADC and Kapp differed between ER-positive and negative tumors (all p < 0.05). Except Kapp, DKI/DWI-parameters showed relation with Ki-67 expression. None of the DKI/DWI-parameters showed relation with lesion grade (all p > 0.05). CONCLUSION: The more complicated and time-consuming DKI is not superior to conventional DWI in differentiating suspicious breast lesions and reflecting prognostic information of breast cancer.

Detection and Classification of Breast Lesions With Readout-Segmented Diffusion-Weighted Imaging in a Large Chinese Cohort.

Objectives: To evaluate the performance of readout-segmented echo-planar imaging DWI (rs-EPI DWI) in detecting and characterizing breast cancers in a large Chinese cohort with comparison to dynamic contrast-enhanced MRI (DCE-MRI). Methods: The institutional review board approved this retrospective study with waived written informed consent. A total of 520 women (mean age, 43.1- ± 10.5-years) were included from July 2013 to October 2019. First, the ability of rs-EPI DWI in detecting breast lesions identified by DCE-MRI was evaluated. The lesion conspicuity of rs-EPI-DWI and DCE-MRI was compared using the Wilcoxon signed rank test. With pathology as a reference, the performance of rs-EPI DWI and DCE-MRI in distinguishing breast cancers was evaluated and compared using the Chi-square test. Results: Of 520 women, 327/520 (62.9%) patients had 423 lesions confirmed by pathology with 203 benign and 220 malignant lesions. The rs-EPI DWI can detect 90.8% (659/726) (reader 1) and 90.6% (663/732) (reader 2) of lesions identified by DCE-MRI. The lesion visibility was superior for DCE-MRI than rs-EPI-DWI (all p < 0.05). With pathology as a reference, the sensitivities and specificities of rs-EPI DWI in diagnosing breast cancers were 95.9% (211/220) and 85.7% (174/203) for reader 1 and 97.7% (215/220) and 86.2% (175/203) for reader 2. No significant differences were found for the performance of DCE-MRI and rs-EPI DWI in discriminating breast cancers (all p > 0.05). Conclusions: Although with an inferior lesion visibility, rs-EPI DWI can detect about 90% of breast lesions identified by DCE-MRI and has comparable diagnostic capacity to that of DCE-MRI in identifying breast cancer.