Explainable hierarchical clustering for patient subtyping and risk prediction.

We present a pipeline in which machine learning techniques are used to automatically identify and evaluate subtypes of hospital patients admitted between 2017 and 2021 in a large UK teaching hospital. Patient clusters are determined using routinely collected hospital data, such as those used in the UK's National Early Warning Score 2 (NEWS2). An iterative, hierarchical clustering process was used to identify the minimum set of relevant features for cluster separation. With the use of state-of-the-art explainability techniques, the identified subtypes are interpreted and assigned clinical meaning, illustrating their robustness. In parallel, clinicians assessed intracluster similarities and intercluster differences of the identified patient subtypes within the context of their clinical knowledge. For each cluster, outcome prediction models were trained and their forecasting ability was illustrated against the NEWS2 of the unclustered patient cohort. These preliminary results suggest that subtype models can outperform the established NEWS2 method, providing improved prediction of patient deterioration. By considering both the computational outputs and clinician-based explanations in patient subtyping, we aim to highlight the mutual benefit of combining machine learning techniques with clinical expertise.

In vivo evaluation of ankle kinematics and tibiotalar joint contact strains using digital volume correlation and 3 T clinical MRI.

BACKGROUND: In vivo evaluation of ankle joint biomechanics is key to investigating the effect of injuries on the mechanics of the joint and evaluating the effectiveness of treatments. The objectives of this study were to 1) investigate the kinematics and contact strains of the ankle joint and 2) to investigate the correlation between the tibiotalar joint contact strains and the prevalence of osteochondral lesions of the talus distribution. METHODS: Eight healthy human ankle joints were subjected to compressive load and 3 T MRIs were obtained before and after applying load. The MR images in combination with digital volume correlation enabled non-invasive measurement of ankle joint kinematics and tibiotalar joint contact strains in three dimensions. FINDINGS: The total translation of the calcaneus was smaller (0.48 ± 0.15 mm, p < 0.05) than the distal tibia (0.93 ± 0.16 mm) and the talus (1.03 ± 0.26 mm). These movements can produce compressive and shear joint contact strains (approaching 9%), which can cause development of lesions on joints. 87.5% of peak tensile, compressive, and shear strains in the tibiotalar joint took place in the medial and lateral zones. INTERPRETATION: The findings suggested that ankle bones translate independently from each other, and in some cases in opposite directions. These findings help explain the distribution of osteochondral lesions of the talus which have previously been observed to be in medial and lateral regions of the talar dome in 90% of cases. They also provide a reason for the central region of talar dome being less susceptible to developing osteochondral lesions.

High resolution three-dimensional strain measurements in human articular cartilage.

An unresolved challenge in osteoarthritis research is characterising the localised intra-tissue mechanical response of articular cartilage. The aim of this study was to explore whether laboratory micro-computed tomography (micro-CT) and digital volume correlation (DVC) permit non-destructive quantification of three-dimensional (3D) strain fields in human articular cartilage. Human articular cartilage specimens were harvested from the knee, mounted into a loading device and imaged in the unloaded and loaded states using a micro-CT scanner. Strain was measured throughout the cartilage volume using the micro-CT image data and DVC analysis. The volumetric DVC-measured strain was within 5% of the known applied strain. Variation in strain distribution between the superficial, middle and deep zones was observed, consistent with the different architecture of the material in these locations. These results indicate DVC method may be suitable for calculating strain in human articular cartilage.

Quantifying 3D Strain in Scaffold Implants for Regenerative Medicine.

Regenerative medicine solutions require thoughtful design to elicit the intended biological response. This includes the biomechanical stimulus to generate an appropriate strain in the scaffold and surrounding tissue to drive cell lineage to the desired tissue. To provide appropriate strain on a local level, new generations of scaffolds often involve anisotropic spatially graded mechanical properties that cannot be characterised with traditional materials testing equipment. Volumetric examination is possible with three-dimensional (3D) imaging, in situ loading and digital volume correlation (DVC). Micro-CT and DVC were utilised in this study on two sizes of 3D-printed inorganic/organic hybrid scaffolds (n = 2 and n = 4) with a repeating homogenous structure intended for cartilage regeneration. Deformation was observed with a spatial resolution of under 200 µm whilst maintaining displacement random errors of 0.97 µm, strain systematic errors of 0.17% and strain random errors of 0.031%. Digital image correlation (DIC) provided an analysis of the external surfaces whilst DVC enabled localised strain concentrations to be examined throughout the full 3D volume. Strain values derived using DVC correlated well against manually calculated ground-truth measurements (R2 = 0.98, n = 8). The technique ensures the full 3D micro-mechanical environment experienced by cells is intimately considered, enabling future studies to further examine scaffold designs for regenerative medicine.

Exploratory Full-Field Mechanical Analysis across the Osteochondral Tissue-Biomaterial Interface in an Ovine Model.

Osteochondral injuries are increasingly prevalent, yet success in articular cartilage regeneration remains elusive, necessitating the development of new surgical interventions and novel medical devices. As part of device development, animal models are an important milestone in illustrating functionality of novel implants. Inspection of the tissue-biomaterial system is vital to understand and predict load-sharing capacity, fixation mechanics and micromotion, none of which are directly captured by traditional post-mortem techniques. This study aims to characterize the localised mechanics of an ex vivo ovine osteochondral tissue-biomaterial system extracted following six weeks in vivo testing, utilising laboratory micro-computed tomography, in situ loading and digital volume correlation. Herein, the full-field displacement and strain distributions were visualised across the interface of the system components, including newly formed tissue. The results from this exploratory study suggest that implant micromotion in respect to the surrounding tissue could be visualised in 3D across multiple loading steps. The methodology provides a non-destructive means to assess device performance holistically, informing device design to improve osteochondral regeneration strategies.

Volatile fluorinated nanoemulsions: a chemical route to controlled delivery of inhalation anesthesia.

Novel dispersions of the volatile inhalation anesthetic sevoflurane have been formulated that can provide controlled, sustainable release of anesthetic over clinically useful timescales. The emulsions can be simply formed with manual shaking, reproducibly yielding droplets of the order of 250 nm diameter, i.e. within the nanoemulsion range. Using a custom flow-rig, release of anesthetic gas from the emulsion has been evaluated, and clinically useful levels achieved through appropriate stirring of the formulation. Stirring can also be used to temporarily increase or decrease the amount of anesthetic released. Once consideration of the unusual nature of the fluorinated systems (phase separation by sedimentation rather than creaming), and the highly perturbed environment of their evaluation (under stirring and flow of gas), the observed behavior regarding sevoflurane evaporation can be reasonably well explained by existing theoretical models. Links between anesthetic release and emulsion structure have been defined, providing the basis for future development.

Potential drug development candidates for human soil-transmitted helminthiases.

BACKGROUND: Few drugs are available for soil-transmitted helminthiasis (STH); the benzimidazoles albendazole and mebendazole are the only drugs being used for preventive chemotherapy as they can be given in one single dose with no weight adjustment. While generally safe and effective in reducing intensity of infection, they are contra-indicated in first-trimester pregnancy and have suboptimal efficacy against Trichuris trichiura. In addition, drug resistance is a threat. It is therefore important to find alternatives. METHODOLOGY: We searched the literature and the animal health marketed products and pipeline for potential drug development candidates. Recently registered veterinary products offer advantages in that they have undergone extensive and rigorous animal testing, thus reducing the risk, cost and time to approval for human trials. For selected compounds, we retrieved and summarised publicly available information (through US Freedom of Information (FoI) statements, European Public Assessment Reports (EPAR) and published literature). Concomitantly, we developed a target product profile (TPP) against which the products were compared. PRINCIPAL FINDINGS: The paper summarizes the general findings including various classes of compounds, and more specific information on two veterinary anthelmintics (monepantel, emodepside) and nitazoxanide, an antiprotozoal drug, compiled from the EMA EPAR and FDA registration files. CONCLUSIONS/SIGNIFICANCE: Few of the compounds already approved for use in human or animal medicine qualify for development track decision. Fast-tracking to approval for human studies may be possible for veterinary compounds like emodepside and monepantel, but additional information remains to be acquired before an informed decision can be made.