Andrea Moschetto

Magnetic Resonance Imaging (MRI) enables the acquisition of multiple image contrasts, such as T1-weighted (T1w) and T2-weighted (T2w) scans, each offering distinct diagnostic insights. However, acquiring all desired modalities increases scan time and cost, motivating research into computational methods for cross-modal synthesis. To address this, recent approaches aim to synthesize missing MRI contrasts from those already acquired, reducing acquisition time while preserving diagnostic quality. Image-to-image (I2I) translation provides a promising framework for this task. In this paper, we present a comprehensive benchmark of generative models–specifically, Generative Adversarial Networks (GANs), diffusion models, and flow matching (FM) techniques–for T1w-to-T2w 2D MRI I2I translation. All frameworks are implemented with comparable settings and evaluated on three publicly available MRI datasets of healthy adults. Our quantitative and qualitative analyses show that the GAN-based Pix2Pix model outperforms diffusion and FM-based methods in terms of structural fidelity, image quality, and computational efficiency. Consistent with existing literature, these results suggest that flow-based models are prone to overfitting on small datasets and simpler tasks, and may require more data to match or surpass GAN performance. These findings offer practical guidance for deploying I2I translation techniques in real-world MRI workflows and highlight promising directions for future research in cross-modal medical image synthesis. Code and models are publicly available at https://github.com/AndreaMoschetto/medical-I2I-benchmark.

Sampled String Matching is presented as an efficient solution to the string matching problem, aiming to tackle the space constraints of indexed string matching while purportedly reducing search times for online solutions. Despite the problem’s inception dating back to 1991, practical solutions have only recently emerged. These purportedly accelerate online searches by up to 35 times compared to conventional methods, achieved through a partial index occupying a mere 5% of the text size. This paper delves into the intricacies of one of the latest and most effective text sampling techniques, character distance sampling, which revolves around sampling distances between characters of a selected alphabet within the text. Specifically, we introduce fake samples while remaining honest! In other words, the study reveals that, interestingly, strategically introducing fake samples within the sampled sequence slashes the required index space by almost half, just avoid compromising the algorithm’s correctness. Additionally, since efficiency is everything, this approach, in turn, purportedly enhances the algorithm’s efficiency under specific conditions.

The exact online string matching problem, pivotal in fields ranging from computational biology to data compression, involves identifying all instances of a specified pattern within a text. Despite extensive examination over the decades, this problem has remained computationally challenging due to the time and space limitations inherent in traditional online and offline methods, respectively. Introduced in 1991, sampled string matching has now emerged as a groundbreaking approach, ingeniously combining classical online string matching techniques with efficient text sampling methods. This approach not only addresses the spatial constraints of indexed string matching but also significantly reduces the search duration in online environments, achieving speed increases of up to hundreds of times while requiring less than 4% of the text size for its partial index. In this paper, we explore the adaptability of various online string matching algorithms within the framework of sampled string matching, which has traditionally relied on the Horspool algorithm. Our investigation reveals that integrating alternative string matching algorithms as subroutines markedly enhances overall performance. These findings highlight the potential for reevaluating established methodologies in light of newer, more dynamic solutions and set the stage for transformative impacts across multiple domains.