Ensemble-based systems are primarily analyzed on how the accuracy of the ensemble depends on that of its members. In this paper, we extend this model with adding a natural constraint regarding a time limit within which the ensemble should make the decision. For this aim, we consider both the execution time and accuracy of each member. Then, we solve the problem on how to find the most accurate ensemble, where the sum of the execution times of its members remains below the limit. As a decision rule, we analyze a majority voting-based one generalized to be applicable in single object detection scenarios. The optimization task leads to a non-separable Knapsack problem, which is addressed using stochastic considerations. The proposed methodology is also validated experimentally for the localization of the optic disc in retinal images.

Breast augmentation was the most commonly performed cosmetic surgery procedure in 2011 in the United States. Although aesthetically pleasing surgical results can only be achieved if the correct breast implant is selected from a large variety of different prosthesis sizes and shapes available on the market, surgeons still rely on visual assessment and other subjective approaches for operative planning because of lacking objective evaluation tools. In this paper, we present the development of a software prototype for augmentation mammaplasty simulation solely based on 3-D surface scans, from which patient-specific finite-element models are generated in a semiautomatic process. The finite-element model is used to preoperatively simulate the expected breast shapes using physical soft-tissue mechanics. Our approach uses a novel mechanism based on so-called displacement templates, which, for a specific implant shape and position, describe the respective internal body forces. Due to a highly efficient numerical solver we can provide immediate visual feedback of the simulation results, and thus, the software prototype can be integrated smoothly into the medical workflow. The clinical value of the developed 3-D computational tool for aesthetic breast augmentation surgery planning is demonstrated in patient-specific use cases.

Generating ensembles from multiple individual classifiers is a popular approach to raise the accuracy of the decision. As a rule for decision making, majority voting is a usually applied model. In this paper, we generalize classical majority voting by incorporating probability terms pn,k to constrain the basic framework. These terms control whether a correct or false decision is made if k correct votes are present among the total number of n. This generalization is motivated by object detection problems, where the members of the ensemble are image processing algorithms giving their votes as pixels in the image domain. In this scenario, the terms pn,k can be specialized by a geometric constraint. Namely, the votes should fall inside a region matching the size and shape of the object to vote together. We give several theoretical results in this new model for both dependent and independent classifiers, whose individual accuracies may also differ. As a real world example, we present our ensemble-based system developed for the detection of the optic disc in retinal images. For this problem, experimental results are shown to demonstrate the characterization capability of this system. We also investigate how the generalized model can help us to improve an ensemble with extending it by adding a new algorithm.

Three-dimensional (3-D) recording of the surface of the human body or anatomical areas has gained importance in many medical specialties. Thus, it is important to determine scanner precision and accuracy in defined medical applications and to establish standards for the recording procedure. Here we evaluated the precision and accuracy of 3-D assessment of the facial area with the Minolta Vivid 910 3D Laser Scanner. We also investigated the influence of factors related to the recording procedure and the processing of scanner data on final results. These factors include lighting, alignment of scanner and object, the examiner, and the software used to convert measurements into virtual images. To assess scanner accuracy, we compared scanner data to those obtained by manual measurements on a dummy. Less than 7% of all results with the scanner method were outside a range of error of 2 mm when compared to corresponding reference measurements. Accuracy, thus, proved to be good enough to satisfy requirements for numerous clinical applications. Moreover, the experiments completed with the dummy yielded valuable information for optimizing recording parameters for best results. Thus, under defined conditions, precision and accuracy of surface models of the human face recorded with the Minolta Vivid 910 3D Scanner presumably can also be enhanced. Future studies will involve verification of our findings using test persons. The current findings indicate that the Minolta Vivid 910 3D Scanner might be used with benefit in medicine when recording the 3-D surface structures of the face.