Search

Domination-based multiobjective (MO) evolutionary algorithms (EAs) are today arguably the most frequently used type of MOEA. These methods, however, stagnate when the majority of the population becomes nondominated, preventing further convergence to the Pareto set. Hypervolume-based MO optimization has shown promising results to overcome this. Direct use of the hypervolume, however, results in no selection pressure for dominated solutions. The recently introduced Sofomore framework overcomes this by solving multiple interleaved single-objective dynamic problems that iteratively improve a single approximation set, based on the uncrowded hypervolume improvement (UHVI). It thereby however loses many advantages of population-based MO optimization, such as handling multimodality. Here, we reformulate the UHVI as a quality measure for approximation sets, called the uncrowded hypervolume (UHV), which can be used to directly solve MO optimization problems with a single-objective optimizer. We use the state-of-the-art gene-pool optimal mixing evolutionary algorithm (GOMEA) that is capable of efficiently exploiting the intrinsically available grey-box properties of this problem. The resulting algorithm, UHV-GOMEA, is compared with Sofomore equipped with GOMEA, and the domination-based MO-GOMEA. In doing so, we investigate in which scenarios either domination-based or hypervolume-based methods are preferred. Finally, we construct a simple hybrid approach that combines MO-GOMEA with UHV-GOMEA and outperforms both. [ABSTRACT FROM AUTHOR]

The Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA) is a model-based EA framework that has been shown to perform well in several domains, including Genetic Programming (GP). Differently from traditional EAs where variation acts blindly, GOMEA learns a model of interdependencies within the genotype, that is, the linkage, to estimate what patterns to propagate. In this article, we study the role of Linkage Learning (LL) performed by GOMEA in Symbolic Regression (SR). We show that the non-uniformity in the distribution of the genotype in GP populations negatively biases LL, and propose a method to correct for this. We also propose approaches to improve LL when ephemeral random constants are used. Furthermore, we adapt a scheme of interleaving runs to alleviate the burden of tuning the population size, a crucial parameter for LL, to SR. We run experiments on 10 real-world datasets, enforcing a strict limitation on solution size, to enable interpretability. We find that the new LL method outperforms the standard one, and that GOMEA outperforms both traditional and semantic GP. We also find that the small solutions evolved by GOMEA are competitive with tuned decision trees, making GOMEA a promising new approach to SR. [ABSTRACT FROM AUTHOR]

Abstract: Background: Functional near-infrared spectroscopy (fNIRS) and functional MRI (fMRI) are non-invasive techniques used to relate activity in different brain regions to certain tasks. Respiratory calibration of the blood oxygen level dependent (BOLD) signal, and combined fNIRS–fMRI approaches have been used to quantify physiological subcomponents giving rise to the BOLD signal. A comparison of absolute oxygen metabolism parameters between MRI and NIRS, using spatially resolved (SRS) NIRS and respiratory calibrated MRI, could yield additional insight in the physiology underlying activation. Materials and methods: Changes in the BOLD signal, cerebral blood flow (CBF), and oxygen saturation (SO2) were derived from a single MRI sequence during a respiratory challenge in healthy volunteers. These changes were compared to SO2 obtained by a single probe SRS NIRS setup. In addition, concentration changes in oxygenated (O2Hb), deoxygenated (HHb), and total haemoglobin (tHb), obtained by NIRS, were compared to the parameters obtained by MRI. Results: NIRS SO2 correlated with end-tidal CO2 (0.83, p<0.0001), the BOLD signal (0.82, p<0.0001), CBF (0.85, p<0.0001), and also MRI SO2 (0.82, p<0.0001). The BOLD signal correlated with NIRS HHb (−0.76, p<0.0001), O2Hb (0.41, p=0.001), and tHb (r=0.32, p=0.01). Conclusions: Good correlations show that changes in cerebral physiology, following a respiratory challenge, go hand in hand with changes in the BOLD signal, CBF, O2Hb, HHb, NIRS SO2, and MRI SO2. Out of all NIRS derived parameters, the SO2 showed the best correlation with the BOLD signal. [Copyright &y& Elsevier]

The article presents a study which investigates the efficacy of marking the axilla with radioactive iodine seeds (MARI) as a new procedure for treating breast cancer. The results show that this procedure, which involves neoadjuvant chemotherapy, is feasible for treatment of breast cancer. It was concluded that this procedure has the potential when making axilla-conserving surgery in treating breast cancer.

Background: For radiotherapy of abdominal cancer, four-dimensional magnetic resonance imaging (4DMRI) is desirable for tumor definition and the assessment of tumor and organ motion. However, irregular breathing gives rise to image artifacts. We developed a outlier rejection strategy resulting in a 4DMRI with reduced image artifacts in the presence of irregular breathing.Methods: We obtained 2D T2-weighted single-shot turbo spin echo images, with an interleaved 1D navigator acquisition to obtain the respiratory signal during free breathing imaging in 2 patients and 12 healthy volunteers. Prior to binning, upper and lower inclusion thresholds were chosen such that 95% of the acquired images were included, while minimizing the distance between the thresholds (inclusion range (IR)). We compared our strategy (Min95) with three commonly applied strategies: phase binning with all images included (Phase), amplitude binning with all images included (MaxIE), and amplitude binning with the thresholds set as the mean end-inhale and mean end-exhale diaphragm positions (MeanIE). We compared 4DMRI quality based on: Data included (DI); percentage of images remaining after outlier rejection. Reconstruction completeness (RC); percentage of bin-slice combinations containing at least one image after binning. Intra-bin variation (IBV); interquartile range of the diaphragm position within the bin-slice combination, averaged over three central slices and ten respiratory bins. IR. Image smoothness (S); quantified by fitting a parabola to the diaphragm profile in a sagittal plane of the reconstructed 4DMRI. A two-sided Wilcoxon's signed-rank test was used to test for significance in differences between the Min95 strategy and the Phase, MaxIE, and MeanIE strategies.Results: Based on the fourteen subjects, the Min95 binning strategy outperformed the other strategies with a mean RC of 95.5%, mean IBV of 1.6 mm, mean IR of 15.1 mm and a mean S of 0.90. The Phase strategy showed a poor mean IBV of 6.2 mm and the MaxIE strategy showed a poor mean RC of 85.6%, resulting in image artifacts (mean S of 0.76). The MeanIE strategy demonstrated a mean DI of 85.6%.Conclusions: Our Min95 reconstruction strategy resulted in a 4DMRI with less artifacts and more precise diaphragm position reconstruction compared to the other strategies.Trial Registration: Volunteers: protocol W15_373#16.007; patients: protocol NL47713.018.14. [ABSTRACT FROM AUTHOR]

Proffered Papers: Proffered papers 12: Artificial Intelligence and automation OC-0225: Highly-individualized dose reconstruction for pediatric abdominal radiotherapy with machine learning M. Virgolin, Z. Wang, B.V. Balgobind, I.W.E.M. Van Dijk, J. Wiersma, D.C. Hodgson, A. Bryce-Atkinson, M. Van Herk, C.R.N. Rasch, L. Zadravec Zaletel, P.S. Kroon, G.O. Janssens, A. Bel, P.A.N. Bosman, T. Alderliesten. [Extracted from the article]