Microcomputed tomography (microCT) is widely used for nondestructive bone phenotyping in small animals, especially in the mouse. Here, we investigated the reproducibility and resolution dependence of microCT analysis of microstructural parameters in three different compartments in the mouse femur. Reproducibility was assessed with respect to precision error (PE%CV) and intraclass correlation coefficient (ICC). We examined 14 left femurs isolated postmortem from two strains of mice (seven per group). Measurements and analyses were repeated five times on different days. In a second step, analysis was repeated again five times for a single measurement. Resolution dependence was assessed by high-resolution measurements (10 microm) in one strain and subsequent image degrading. Reproducibility was better in full bone compartment and in cortical bone compartment in the diaphysis (PE%CV = 0.06-2.16%) than in trabecular compartment in the distal metaphysis (PE(%CV) = 0.59-5.24%). Nevertheless, ICC (0.92-1.00) showed a very high reliability of the assessed parameters in all regions, indicating very small variances within repeated measurements compared to the population variances. Morphometric indices computed from lower- and higher-resolution images displayed in general only weak dependence and were highly correlated with each other (R2 = 0.91-0.99). The results show that parameters in the full and cortical compartments were very reproducible, whereas precision in the trabecular compartment was somewhat lower. Nevertheless, all compartmental analysis methods were very robust, as shown by the high ICC values, demonstrating high suitability for application in inbred strains, where highest precision is needed due to small population variances.

No abstract

We study the design of iterative combinatorial auctions for domains with a large number of items. In such domains, preference elicitation is a major challenge because the bundle space grows exponentially in the number of items. To keep preference elicitation manageable, recent work has employed a machine learning (ML) algorithm that identifies a small set of bundles to query from each bidder. However, a major limitation of this prior work is that bidders must submit exact values for the queried bundles, which can be quite costly for them. To address this, we propose a new ML-powered auction with interval bidding (i.e., where bidders submit upper and lower bounds for the queried bundles). To steer the auction towards an efficient allocation, we design a price-based refinement process asking bidders to tighten bounds on relevant bundles only, and we carefully integrate this refinement process into the ML-based query module. Our experiments show that our new auction with interval bidding achieves almost the same allocative efficiency as the prior auction design that required bidders to submit exact values. Despite only eliciting interval bids, our auction beats the well-known combinatorial clock auction in a realistically-sized domain.

No abstract