Multimodal Artificial Intelligence in Medicine.

Traditional medical Artificial Intelligence models, approved for clinical use, restrict themselves to single-modal data e.g. images only, limiting their applicability in the complex, multimodal environment of medical diagnosis and treatment. Multimodal Transformer Models in healthcare can effectively process and interpret diverse data forms such as text, images, and structured data. They have demonstrated impressive performance on standard benchmarks like USLME question banks and continue to improve with scale. However, the adoption of these advanced AI models is not without challenges. While multimodal deep learning models like Transformers offer promising advancements in healthcare, their integration requires careful consideration of the accompanying ethical and environmental challenges.

Dielectric characterisation of human adrenal glands and adrenal tumours for the development of microwave ablation technologies for hypertension treatment.

Adrenal gland-induced hypertension results from underlying adrenal gland disorders including Conn's syndrome, Cushing's syndrome, and Pheochromocytoma. These adrenal disorders are a risk for cardiovascular and renal morbidity and mortality. Clinically, treatment for adrenal gland-induced hypertension involves a pharmaceutical or surgical approach. The former presents very significant side effects whereas the latter can be ineffective in cases where the adrenal disorder reoccurs in the remaining contralateral adrenal gland. Due to the limitations of existing treatment methods, minimally invasive treatment options like microwave ablation (MWA) have received significant attention for treating adrenal gland disorders. A precise comprehension of the dielectric properties of human adrenal glands will help to tailor energy delivery for MWA therapy, thus offering the potential to optimise treatments and minimise damage to surrounding tissues. This study reports the ex vivo dielectric properties of human adrenal glands, including the cortex, medulla, capsule, and tumours, based on the data obtained from four patients (diagnosed with Conn's syndrome, Cushing's syndrome, and Pheochromocytoma) who underwent unilateral adrenalectomy at the University Hospital Galway, Ireland. An open-ended coaxial probe measurement technique was used to measure the dielectric properties for a frequency range of 0.5-8.5 GHz. The dielectric properties were fitted using a two-pole Debye model, and a weighted least squares method was employed to optimise the model parameters. Moreover, the dielectric properties of adrenal tissues and tumours were compared across frequencies commonly used in MWA, including 915 MHz, 2.45 GHz, and 5.8 GHz. The study found that the dielectric properties of adrenal tumours were influenced by the presence of lipid-rich adenomas, and the dielectric properties of Cushing's syndrome tumour were lowest in comparison to the tumours in patients diagnosed with Conn's syndrome and Pheochromocytoma. Furthermore, a notable difference was observed in the dielectric properties of the medulla and cortex among patients diagnosed with Conn's syndrome, Cushing's syndrome, and Pheochromocytoma. These findings have significant implications for the diagnosis and treatment of adrenal tumours, including the optimisation of MWA therapy for precise ablation of adrenal masses.

Web-Based Interfaces for Virtual C. elegans Neuron Model Definition, Network Configuration, Behavioral Experiment Definition and Experiment Results Visualization.

The Si elegans platform targets the complete virtualization of the nematode Caenorhabditis elegans, and its environment. This paper presents a suite of unified web-based Graphical User Interfaces (GUIs) as the main user interaction point, and discusses their underlying technologies and methods. The user-friendly features of this tool suite enable users to graphically create neuron and network models, and behavioral experiments, without requiring knowledge of domain-specific computer-science tools. The framework furthermore allows the graphical visualization of all simulation results using a worm locomotion and neural activity viewer. Models, experiment definitions and results can be exported in a machine-readable format, thereby facilitating reproducible and cross-platform execution of in silico C. elegans experiments in other simulation environments. This is made possible by a novel XML-based behavioral experiment definition encoding format, a NeuroML XML-based model generation and network configuration description language, and their associated GUIs. User survey data confirms the platform usability and functionality, and provides insights into future directions for web-based simulation GUIs of C. elegans and other living organisms. The tool suite is available online to the scientific community and its source code has been made available.

Optimised analytical models of the dielectric properties of biological tissue.

The interaction of electromagnetic fields with the human body is quantified by the dielectric properties of biological tissues. These properties are incorporated into complex numerical simulations using parametric models such as Debye and Cole-Cole, for the computational investigation of electromagnetic wave propagation within the body. These parameters can be acquired through a variety of optimisation algorithms to achieve an accurate fit to measured data sets. A number of different optimisation techniques have been proposed, but these are often limited by the requirement for initial value estimations or by the large overall error (often up to several percentage points). In this work, a novel two-stage genetic algorithm proposed by the authors is applied to optimise the multi-pole Debye parameters for 54 types of human tissues. The performance of the two-stage genetic algorithm has been examined through a comparison with five other existing algorithms. The experimental results demonstrate that the two-stage genetic algorithm produces an accurate fit to a range of experimental data and efficiently out-performs all other optimisation algorithms under consideration. Accurate values of the three-pole Debye models for 54 types of human tissues, over 500 MHz to 20 GHz, are also presented for reference.