Prof Patrick Guidotti (Department of Mathematics, University of California, Irvine) 

Connecting the Dots: Extracting Manifold information from Point Clouds.

We describe a method to produce geometric information from a point cloud that has an analytical justification when the point cloud is the sample of a smooth manifold. The method is in the spirit of numerical mesh free methods and allows for the recovery of the normal and the curvatures when the point cloud is the sample of a hyper-surface. Based on this information one has access to the surface nabla and to its Laplace-Beltrami operator directly from a set of values on the point cloud. We also make a connection between kernel interpolation and Gaussian Process regression that yields a justification for use of approximate kernel based interpolation (a regularised version of interpolation). The latter proves very useful when dealing with noisy data, i.e. when the point clouds or the values associated to it are polluted by noise. Several numerical experiments will be shown that highlight the effectiveness of the method.

Prof Huan Han (Department of Mathematics, Wuhan University of Technology)

Medical image registration: variational framework and few shot learning.

Physical mesh folding elimination is a key challenge in medical image registration. This arise the so called diffeomorphic image registration who searches for a C1 1-to-1 mapping/diffeomorphism. It is essentially a PDE/inequality constrained variational problem. For the research on this topic, it is faced up with two challenges: 1. How to give a simple constraint to ensure the mapping to be diffeomorphism; 2. No labels for training the deep learning network; 3. Lack of large amount of image. Surrounding the above three challenges, this talk establishes a conformal variational framework for diffeomorphic image registration, discuss the asymptotic behaviour of solution with respect to the hyper-parameters and related few-shot learning network. Numerical examples are demonstrated to validate the efficiency of the proposed framework. This is a joint work with Ke Chen, Junping Wei, Zhengping Wang and Yimin Zhang.

Dr Matthew Bourn (National Measurement Laboratory at LGC, UK)

Integrating AI into Cell Characterisation: A Metrological Perspective. 

Measurement of cells often requires a multidisciplinary approach to account for associated complexities. Wider information on critical quality attributes (CQAs) such as cell count, viability and functionality improve the accuracy of cell analysis whether this is for diagnostics, cell therapy, disease monitoring, or fundamental research. Defining CQAs for a cell product remains challenging due to the biological complexity and lack of standardisation in existing flow cytometric methodology. This results in considerable irreproducibility which confounds an already challenging measurement space. The National Measurement Institutes (NMIs) globally invest a concerted effort to create measurement assurance methods and underpin measurement systems with traceable standards through interlaboratory comparison studies and development of new ISO standards for cell characterisation. As cutting-edge AI approaches become increasingly embedded in high-throughput cell analysis workflows, ensuring the accuracy, reliability, and reproducibility of results is more critical than ever. As artificial intelligence becomes an integral component of high-throughput cell analysis, it is imperative to address the metrological implications of its integration. This presentation will examine the measurement uncertainties introduced by AI-driven morphological profiling in microscopy and flow cytometry, highlighting the need for rigorous validation to ensure data accuracy and robustness.

Prof Ken Chen (Department of Mathematics and Statistics, University of Strathclyde)

Can novel mathematical models outperform AI for image segmentation? 

While AI models represent both a significant advancement and a promising future direction in image processing, it remains unclear whether AI alone can handle the most complex cases. Tasks such as image segmentation are routinely addressed by existing models; however, certain particularly difficult images remain unsolved due to their inherent complexity. In contrast, mathematical models based on level set formulations rely on appropriate regularization to succeed. For 1D and 2D images, effective regularizes include mean curvature, Gaussian curvature, and Euler’s elastica. While these have been extended to 3D, the resulting approaches often exhibit theoretical or computational shortcomings. A more principled extension to 3D involves Ricci tensors and normal curvature. This work explores this direction, presenting both the theoretical framework and algorithmic development of a variational segmentation model. Experimental results demonstrate that the proposed model outperforms state-of-the-art mathematical and AI-based methods in 3D image segmentation. Joint work with Jisui Huang (CNU), Andreas Alpers (Liverpool) and Na Lei (Dalian).

Prof Tieyong Zeng (Centre for Mathematical Artificial Intelligence, Chinese University of Hong Kong)

High Frequency Modulated Transformer for Multi-Contrast MRI Super-Resolution.

Accelerating the MRI acquisition process is always a key issue in modern medical practice, and great efforts have been devoted to fast MR imaging. Among them, multi-contrast MR imaging is a promising and effective solution that utilizes and combines information from different contrasts. However, existing methods may ignore the importance of the high-frequency priors among different contrasts. Moreover, they may lack an efficient method to fully utilize the information from the reference contrast. In this talk, we propose a lightweight and accurate High-frequency Modulated Transformer (HFMT) for multi-contrast MRI super-resolution. The key ideas of HFMT are high-frequency prior enhancement and its fusion with global features. Specifically, we employ an enhancement module to enhance and amplify the high-frequency priors in the reference and target modalities. In addition, we utilize the Rectangle Window Transformer Block (RWTB) to capture global information in the target contrast. Meanwhile, we propose a novel cross-attention mechanism to fuse the high-frequency enhanced features with the global features sequentially, which assists the network in recovering clear texture details from the low-resolution inputs. Extensive experiments show that our proposed method can reconstruct high-quality images with fewer parameters and faster inference time.

Dr Nicholas Rattray (SIPBS, University of Strathclyde)

Predictive Biomarkers of Surgical Outcomes in older Adults Through Metabolic Profiling.

The ageing population is rapidly becoming one of the most critical global public health challenges. The World Health Organization projects that by 2050, the number of adults aged 60 and over will surpass 2 billion—more than triple the figure from 2000. Ageing is marked by declining physiological resilience, compromised biomolecular functions, and increased vulnerability to death. Tackling these changes demands new biochemical tools that enable earlier detection and clinical intervention.

Our research employs a systems biogerontology approach that integrates mass spectrometry–based metabolic modeling, Mendelian randomization genetics, and gene set enrichment to help develop understanding of biomolecular mechanisms of ageing and frailty. Our strategy has previously identified 12 dysregulated metabolites linked to frailty processes, including those in glycolysis, the carnitine shuttle, and Vitamin E metabolism. These metabolic signatures emerged from population-level discovery work and are being refined for clinical relevance.

In a recent surgical trial on clinical frailty we have collected blood plasma samples, generated a range of biomolecular datasets and applied generalised linear models (GLMs) to assess metabolite-age associations while adjusting for confounders. LASSO regression has also been used to identify robust, minimal biomarker panels by filtering out low-impact variables in high-dimensional data. The application of PC-CVA also enabled clear visualizations of metabolic differences between time-course data, aiding the interpretation of group-level biochemical variation. Overall out methods have developed a robust and predictive biomarkers of post-operative surgical outcomes in older adults and lays the groundwork for translating metabolic frailty research into personalized and preventive healthcare strategies.

Dr Zahra Rattray (SIPBS, University of Strathclyde)

Improving Injectability of mAb Formulations via Computational and Experimental Surface Optimization. 

High viscosity in high-concentration monoclonal antibody (mAb) formulations poses critical challenges to manufacturability, product quality, and immunogenicity. With the increasing adoption of subcutaneous autoinjectors—favoured for enhancing patient adherence and reducing healthcare visits—viscosity becomes a key limiting factor. These devices typically accommodate ~1 mL volumes, and high viscosity protein solutions often result in device failure due to the elevated force required for full-dose delivery.

We employed in silico surface patch analysis and molecular descriptor calculations using MOE to design a panel of IgG1 variants with single-point mutations in the variable region. These mutations targeted solvent-accessible charged or hydrophobic surface patches. Comprehensive biophysical characterization was conducted to assess parameters such as viscosity, net hydrophobicity, antigen affinity, and thermal stability.

Single-point mutations induced notable changes in surface patch distribution, molecular charge, hydrophobicity, and self-association propensity. Viscosity reductions were primarily driven by hydrophobic interactions. However, both computational predictions and experimental measurements poorly correlated with viscosity.

Developability assessments are vital for progressing mAb candidates into clinical development. Sequence optimization via targeted mutations is a promising strategy to enhance biophysical properties. While hydrophobicity-driven self-interactions are key to viscosity, our findings highlight the need for case-by-case evaluation due to inherent variability in mAb surface potential landscapes.