Traditional information retrieval systems typically rely on a single attribute to rank items, often neglecting the complex relationships that may exist among item attributes. As a result, users frequently need to refine their queries to obtain relevant results. To address this limitation, the Composite Retrieval (CR) framework has been proposed, which aims to construct diverse and complementary bundles of items. The goal is to maximise intra-bundle similarity while simultaneously enhancing inter-bundle complementarity, thereby providing more meaningful results without requiring further user intervention. This work introduces a novel bi-objective formulation for the Composite Retrieval Problem that explicitly models the trade-off between bundle cohesion and diversity. To solve this problem, we implement a variant of the NSGA-II algorithm using the jMetal framework, enabling the generation of a set of Pareto-optimal solutions.

This paper tackles the multi-objective Direct Aperture Optimisation (DAO) problem in Intensity-Modulated Radiation Therapy (IMRT) by introducing a Pareto Local Search (PLS) heuristic. Unlike traditional single-objective methods, the proposed approach captures the trade-offs between tumour coverage and organ-at-risk (OAR) sparing, reflecting the multi-objective nature of clinical planning. The algorithm is tested on a prostate cancer case using the CERR platform and compared against a weighted sum local search baseline. Experimental results show that PLS produces a superior hypervolume performance, offering a more accurate approximation of the Pareto front. These results highlight the potential of Pareto-based heuristics to improve the quality and clinical relevance of IMRT treatment plans.

The planning of Intensity-Modulated Radiation Therapy (IMRT) treatments is highly sensitive to patient-specific anatomical variations and machine setup errors. These uncertainties can significantly degrade the quality of the delivered dose distribution, leading to suboptimal clinical outcomes. In this work, we propose a mono-objective robust optimisation approach for the Beam Angle Optimisation (BAO) problem, focusing on mitigating the effects of anatomical variability. A scenario-based model is developed to simulate uncertainties associated with setup errors and anatomical displacements. A Local Search algorithm is implemented to iteratively refine beam configurations, minimising a penalised objective function across worst-case scenarios. Experimental results on clinical datasets demonstrate the effectiveness of the proposed method in improving plan robustness while maintaining computational efficiency. The methodology shows promise for enhancing the reliability of IMRT treatments without substantially increasing planning complexity.

Intensity-modulated radiation therapy (IMRT) is one of the most common radiation techniques used in cancer treatment. The main goal of IMRT is to deliver the prescribed radiation dose to the tumour area while sparing surrounding healthy organs. Traditionally, the IMRT problem has been divided into three sequential sub-problems. First, the beam angles from which radiation is delivered are selected. This problem is called the beam angle optimisation (BAO) problem. Then, the optimal radiation intensity for the beam angles selected in the previous step is computed in the so-called fluence map optimisation problem (FMO). Finally, the aperture shapes and their sequencing are generated by solving a sequencing problem (MLC). One drawback of this is that each step's outcome heavily depends on the quality of the previous step. Further, the optimal intensities obtained during the FMO problem are usually impossible to deliver using the available physical devices. Thus, the treatment plan computed during the second step of the sequential approach suffers an important impairment during the aperture shape optimisation process. To address this issue, some researchers have integrated the FMO problem and the MLC into a problem called the direct aperture optimisation (DAO) problem. In the DAO problem, the idea is that, given a set of beam angles, it seeks a set of aperture shapes and their corresponding intensities. Obtaining a treatment plan is fully deliverable, unlike in the FMO problem. However, the DAO problem still depends on the quality of the beam angles selected in the BAO problem. In this study, we propose an integrated BAO + DAO that, to our knowledge, has not been previously proposed in the literature. We solve the integrated problem by using a Local Search to seek high-quality beam angle configurations and a local search to solve that DAO problem for each generated beam angle configuration.