Clinical machine learning faces a critical dilemma in high-stakes medical applications: algorithms achieving optimal diagnostic performance typically sacrifice the interpretability essential for physician decision-making, while interpretable methods compromise sensitivity in complex scenarios. This paradox becomes particularly acute in non-invasive prenatal testing (NIPT), where missed chromosomal abnormalities carry profound clinical consequences yet regulatory frameworks mandate explainable AI systems. We introduce Medical Priority Fusion (MPF), a constrained multi-objective optimization framework that resolves this fundamental trade-off by systematically integrating Naive Bayes probabilistic reasoning with Decision Tree rule-based logic through mathematically-principled weighted fusion under explicit medical constraints. Rigorous validation on 1,687 real-world NIPT samples characterized by extreme class imbalance (43.4:1 normal-to-abnormal ratio) employed stratified 5-fold cross-validation with comprehensive ablation studies and statistical hypothesis testing using McNemar's paired comparisons. MPF achieved simultaneous optimization of dual objectives: 89.3% sensitivity (95% CI: 83.9-94.7%) with 80% interpretability score, significantly outperforming individual algorithms (McNemar's test, p < 0.001). The optimal fusion configuration achieved Grade A clinical deployment criteria with large effect size (d = 1.24), establishing the first clinically-deployable solution that maintains both diagnostic accuracy and decision transparency essential for prenatal care. This work demonstrates that medical-constrained algorithm fusion can resolve the interpretability-performance trade-off, providing a mathematical framework for developing high-stakes medical decision support systems that meet both clinical efficacy and explainability requirements.

The application of information theory in medical imaging, particularly in magnetic resonance imaging (MRI), offers powerful quantitative tools for analyzing structural differences in brain tissues. This study utilizes mutual information (MI) and pixel intensity distributions to differentiate between normal and tumor-affected brain MRI images. Mutual information analyses revealed significantly higher MI values in tumor images compared to normal ones, indicating greater internal similarity within tumor images. Pixel intensity analysis further demonstrated distinct distribution patterns between the two groups: tumor images showed pronounced pixel frequency concentrations within a specific intensity range (0.3, 0.4), suggesting predictable structural characteristics. Conversely, normal images exhibited broader, more uniform pixel intensity distributions across most intensity ranges, except for an initial peak observed in both groups. These findings highlight the capability of information-theoretic metrics, such as mutual information and pixel intensity analysis, to effectively distinguish tumor tissue from normal brain structures, providing promising avenues for enhanced diagnostic and analytical methods in neuroimaging.

This study proposes a deep learning framework and annotation methodology for the automatic detection of periodontal bone loss landmarks, associated conditions, and staging. 192 periapical radiographs were collected and annotated with a stage agnostic methodology, labelling clinically relevant landmarks regardless of disease presence or extent. We propose a heuristic post-processing module that aligns predicted keypoints to tooth boundaries using an auxiliary instance segmentation model. An evaluation metric, Percentage of Relative Correct Keypoints (PRCK), is proposed to capture keypoint performance in dental imaging domains. Four donor pose estimation models were adapted with fine-tuning for our keypoint problem. Post-processing improved fine-grained localisation, raising average PRCK^{0.05} by +0.028, but reduced coarse performance for PRCK^{0.25} by -0.0523 and PRCK^{0.5} by -0.0345. Orientation estimation shows excellent performance for auxiliary segmentation when filtered with either stage 1 object detection model. Periodontal staging was detected sufficiently, with the best mesial and distal Dice scores of 0.508 and 0.489, while furcation involvement and widened periodontal ligament space tasks remained challenging due to scarce positive samples. Scalability is implied with similar validation and external set performance. The annotation methodology enables stage agnostic training with balanced representation across disease severities for some detection tasks. The PRCK metric provides a domain-specific alternative to generic pose metrics, while the heuristic post-processing module consistently corrected implausible predictions with occasional catastrophic failures. The proposed framework demonstrates the feasibility of clinically interpretable periodontal bone loss assessment, with potential to reduce diagnostic variability and clinician workload.

Colorectal cancer (CRC) poses a major public health challenge due to its increasing prevalence, particularly among younger populations. Microsatellite instability-high (MSI-H) CRC and deficient mismatch repair (dMMR) CRC constitute 15% of all CRC and exhibit remarkable responsiveness to immunotherapy, especially with PD-1 inhibitors. Despite this, there is a significant need to optimise immunotherapeutic regimens to maximise clinical efficacy and patient quality of life whilst minimising monetary costs. To address this, we employ a novel framework driven by delay integro-differential equations to model the interactions among cancer cells, immune cells, and immune checkpoints. Several of these components are being modelled deterministically for the first time in cancer, paving the way for a deeper understanding of the complex underlying immune dynamics. We consider two compartments: the tumour site and the tumour-draining lymph node, incorporating phenomena such as dendritic cell (DC) migration, T cell proliferation, and CD8+ T cell exhaustion and reinvigoration. Parameter values and initial conditions are derived from experimental data, integrating various pharmacokinetic, bioanalytical, and radiographic studies, along with deconvolution of bulk RNA-sequencing data from the TCGA COADREAD and GSE26571 datasets. We finally optimised neoadjuvant treatment with pembrolizumab, a widely used PD-1 inhibitor, to balance efficacy, efficiency, and toxicity in locally advanced MSI-H/dMMR CRC patients. We mechanistically analysed factors influencing treatment success and improved upon currently FDA-approved therapeutic regimens for metastatic MSI-H/dMMR CRC, demonstrating that a single medium-to-high dose of pembrolizumab may be sufficient for effective tumour eradication while being efficient, safe and practical.

The Circle of Willis (CoW) is an important network of arteries connecting major circulations of the brain. Its vascular architecture is believed to affect the risk, severity, and clinical outcome of serious neurovascular diseases. However, characterizing the highly variable CoW anatomy is still a manual and time-consuming expert task. The CoW is usually imaged by two non-invasive angiographic imaging modalities, magnetic resonance angiography (MRA) and computed tomography angiography (CTA), but there exist limited datasets with annotations on CoW anatomy, especially for CTA. Therefore, we organized the TopCoW challenge with the release of an annotated CoW dataset. The TopCoW dataset is the first public dataset with voxel-level annotations for 13 CoW vessel components, enabled by virtual reality technology. It is also the first large dataset using 200 pairs of MRA and CTA from the same patients. As part of the benchmark, we invited submissions worldwide and attracted over 250 registered participants from six continents. The submissions were evaluated on both internal and external test datasets of 226 scans from over five centers. The top performing teams achieved over 90% Dice scores at segmenting the CoW components, over 80% F1 scores at detecting key CoW components, and over 70% balanced accuracy at classifying CoW variants for nearly all test sets. The best algorithms also showed clinical potential in classifying fetal-type posterior cerebral artery and locating aneurysms with CoW anatomy. TopCoW demonstrated the utility and versatility of CoW segmentation algorithms for a wide range of downstream clinical applications with explainability. The annotated datasets and best performing algorithms have been released as public Zenodo records to foster further methodological development and clinical tool building.

Cytasters have been underestimated in terms of their potential relevance to embryonic development and evolution. From the perspective discussed herein, structures such as the multiciliated cells of comb rows and balancers supporting mineralized statoliths and macrocilia in Beroe ovata point to a past event of multiflagellate fusion in the origin of metazoans. These structures, which are unique in evolutionary history, indicate that early animals handled basal bodies and their duplication in a manner consistent with a "developmental program" originated in the Ctenophora. Furthermore, the fact that centrosome amplification leads to spontaneous tumorigenesis suggests that the centrosome regulation process was co-opted into a neoplastic functional module. Multicilia, cilia, and flagella are deeply rooted in the evolution of animals and Neoplasia. The fusion of several flagellated microgametes into a cell with a subsequent phase of zygotic (haplontic) meiosis might have been at the origin of both animal evolution and the neoplastic process. In the Ediacaran ocean, we also encounter evolutionary links between the Warburg effect and Neoplasia.

Center-based models are used to simulate the mechanical behavior of biological cells during embryonic development or cancer growth. To allow for the simulation of biological populations potentially growing from a few individual cells to many thousands or more, these models have to be numerically efficient, while being reasonably accurate on the level of individual cell trajectories. In this work, we increase the robustness, accuracy, and efficiency of the simulation of center-based models by choosing the time steps adaptively in the numerical method. We investigate the gain in using single rate time stepping for the forward and backward Euler methods, based on local estimates of the numerical errors and the stability of the method in the case of the explicit forward Euler method. Furthermore, we propose a multirate time stepping scheme that simulates regions with high local force gradients (e.g. as they happen after cell division) with multiple smaller time steps within a larger single time step for regions with smoother forces. These methods are compared for different model systems in numerical experiments. We conclude that the adaptive single rate forward Euler method results in significant gains in terms of reduced wall clock times for the simulation of a linearly growing tissue, while at the same time eliminating the need for manual determination of a suitable time step size.

The use of artificial intelligence to enable precision medicine and decision support systems through the analysis of pathology images has the potential to revolutionize the diagnosis and treatment of cancer. Such applications will depend on models' abilities to capture the diverse patterns observed in pathology images. To address this challenge, we present Virchow, a foundation model for computational pathology. Using self-supervised learning empowered by the DINOv2 algorithm, Virchow is a vision transformer model with 632 million parameters trained on 1.5 million hematoxylin and eosin stained whole slide images from diverse tissue and specimen types, which is orders of magnitude more data than previous works. The Virchow model enables the development of a pan-cancer detection system with 0.949 overall specimen-level AUC across 17 different cancer types, while also achieving 0.937 AUC on 7 rare cancer types. The Virchow model sets the state-of-the-art on the internal and external image tile level benchmarks and slide level biomarker prediction tasks. The gains in performance highlight the importance of training on massive pathology image datasets, suggesting scaling up the data and network architecture can improve the accuracy for many high-impact computational pathology applications where limited amounts of training data are available.

We aimed to develop a population pharmacokinetic model of tacrolimus in Chinese lung transplant recipients, and propose model based dosing regimens for individualized treatment. We obtained 807 tacrolimus whole blood concentrations from 52 lung transplant patients and genotyped CYP3A5*3. Population pharmacokinetic analysis was performed using nonlinear mixed effects modeling. Monte Carlo simulations were employed to design initial dosing regimens. Tacrolimus pharmacokinetics was described by a one compartment model with first order absorption and elimination process. The mean estimated apparent clearance was 13.1 l/h with 20.1% inter subject variability in CYP3A5*3/*3 70kg patients with 30% hematocrit and voriconazole free therapy, which is lower than that in Caucasian(17.5 to 36.5 l/h). Hematocrit, postoperative days, tacrolimus daily dose, voriconazole cotherapy, and CYP3A5*3 genotype were identified as significant covariates for tacrolimus clearance. To achieve the target trough concentration (10 to 15 ng/ml) on the 8th day after transplantation, CYP3A5*1/*3 patients with voriconazole free cotherapy, a higher initial dosage than the current regimen of 0.04 mg/kg q12h should be recommened. Given the nonlinear kinetics of tacrolimus and large variability, population pharmacokinetic model should be combined with therapeutic drug monitoring to optimize individualized therapy.

Cancer prognostication is a challenging task in computational pathology that requires context-aware representations of histology features to adequately infer patient survival. Despite the advancements made in weakly-supervised deep learning, many approaches are not context-aware and are unable to model important morphological feature interactions between cell identities and tissue types that are prognostic for patient survival. In this work, we present Patch-GCN, a context-aware, spatially-resolved patch-based graph convolutional network that hierarchically aggregates instance-level histology features to model local- and global-level topological structures in the tumor microenvironment. We validate Patch-GCN with 4,370 gigapixel WSIs across five different cancer types from the Cancer Genome Atlas (TCGA), and demonstrate that Patch-GCN outperforms all prior weakly-supervised approaches by 3.58-9.46%. Our code and corresponding models are publicly available at https://github.com/mahmoodlab/Patch-GCN.

Collective cell migration is a key driver of embryonic development, wound healing, and some types of cancer invasion. Here we provide a physical perspective of the mechanisms underlying collective cell migration. We begin with a catalogue of the cell-cell and cell-substrate interactions that govern cell migration, which we classify into positional and orientational interactions. We then review the physical models that have been developed to explain how these interactions give rise to collective cellular movement. These models span the sub-cellular to the supracellular scales, and they include lattice models, phase fields models, active network models, particle models, and continuum models. For each type of model, we discuss its formulation, its limitations, and the main emergent phenomena that it has successfully explained. These phenomena include flocking and fluid-solid transitions, as well as wetting, fingering, and mechanical waves in spreading epithelial monolayers. We close by outlining remaining challenges and future directions in the physics of collective cell migration.

The formation of new bone involves both the deposition of bone matrix, and the formation of a network of cells embedded within the bone matrix, called osteocytes. Osteocytes derive from bone-synthesising cells (osteoblasts) that become buried in bone matrix during bone deposition. The generation of osteocytes is a complex process that remains incompletely understood. Whilst osteoblast burial determines the density of osteocytes, the expanding network of osteocytes regulates in turn osteoblast activity and osteoblast burial. In this paper, a spatiotemporal continuous model is proposed to investigate the osteoblast-to-osteocyte transition. The aims of the model are (i) to link dynamic properties of osteocyte generation with properties of the osteocyte network imprinted in bone, and (ii) to investigate Marotti's hypothesis that osteocytes prompt the burial of osteoblasts when they become covered with sufficient bone matrix. Osteocyte density is assumed in the model to be generated at the moving bone surface by a combination of osteoblast density, matrix secretory rate, rate of entrapment, and curvature of the bone substrate, but is found to be determined solely by the ratio of the instantaneous burial rate and matrix secretory rate. Osteocyte density does not explicitly depend on osteoblast density nor curvature. Osteocyte apoptosis is also included to distinguish between the density of osteocyte lacuna and the density of live osteocytes. Experimental measurements of osteocyte lacuna densities are used to estimate the rate of burial of osteoblasts in bone matrix. These results suggest that: (i) burial rate decreases during osteonal infilling, and (ii) the control of osteoblast burial by osteocytes is likely to emanate as a collective signal from a large group of osteocytes, rather than from the osteocytes closest to the bone deposition front.

How genes affect tissue scale organization remains a longstanding biological puzzle. As experimental efforts aim to quantify gene expression, chromatin organization, cellular structure, and tissue structure, computational modeling lags behind. To address this gap, we merge a cellular-based tissue model with a nuclear model that includes a deformable lamina shell and chromatin to test multiscale hypotheses linking chromatin and tissue scales. We propose a multiscale hypothesis focusing on brain organoids to explain structural differences between brain organoids built from induced-pluripotent human stem cells and induced-pluripotent gorilla and chimpanzee cells. Recent experiments discover that a cell fate transition from neuroepithelial to radial glial cells includes a new intermediate state delayed in human organoids, which narrows and lengthens cells on the apical side. Experiments show that the transcription factor ZEB2 plays a major role in the emergence of this intermediate state with ZEB2 mRNA levels peaking. We postulate that the enhancement of ZEB2 expression is potentially due to chromatin reorganization in response to mechanical deformations of the nucleus. A larger critical mechanical strain triggers reorganization in human-derived stem cells, causing delayed ZEB2 upregulation compared with genetically close relatives. We test this by exploring how slightly different initial configurations of chromatin reorganize under applied strain, with greater representing less genetically-close relatives. We find that larger configuration discrepancies produce increased differences in the magnitude of chromatin displacement that rise faster than linearly yet slower than exponentially. Changes in chromatin strain and contact maps can reveal species-specific differences, aiding our understanding of how one species differs in structure from another.

The interactions between cells and the extracellular matrix are vital for the self-organisation of tissues. In this paper we present proof-of-concept to use machine learning tools to predict the role of this mechanobiology in the self-organisation of cell-laden hydrogels grown in tethered moulds. We develop a process for the automated generation of mould designs with and without key symmetries. We create a large training set with $N=6500$ cases by running detailed biophysical simulations of cell-matrix interactions using the contractile network dipole orientation (CONDOR) model for the self-organisation of cellular hydrogels within these moulds. These are used to train an implementation of the \texttt{pix2pix} deep learning model, reserving $740$ cases that were unseen in the training of the neural network for training and validation. Comparison between the predictions of the machine learning technique and the reserved predictions from the biophysical algorithm show that the machine learning algorithm makes excellent predictions. The machine learning algorithm is significantly faster than the biophysical method, opening the possibility of very high throughput rational design of moulds for pharmaceutical testing, regenerative medicine and fundamental studies of biology. Future extensions for scaffolds and 3D bioprinting will open additional applications.

Recent advances in synthetic imaging open up opportunities for obtaining additional data in the field of surgical imaging. This data can provide reliable supplements supporting surgical applications and decision-making through computer vision. Particularly the field of image-guided surgery, such as laparoscopic and robotic-assisted surgery, benefits strongly from synthetic image datasets and virtual surgical training methods. Our study presents an intuitive approach for generating synthetic laparoscopic images from short text prompts using diffusion-based generative models. We demonstrate the usage of state-of-the-art text-to-image architectures in the context of laparoscopic imaging with regard to the surgical removal of the gallbladder as an example. Results on fidelity and diversity demonstrate that diffusion-based models can acquire knowledge about the style and semantics in the field of image-guided surgery. A validation study with a human assessment survey underlines the realistic nature of our synthetic data, as medical personnel detects actual images in a pool with generated images causing a false-positive rate of 66%. In addition, the investigation of a state-of-the-art machine learning model to recognize surgical actions indicates enhanced results when trained with additional generated images of up to 5.20%. Overall, the achieved image quality contributes to the usage of computer-generated images in surgical applications and enhances its path to maturity.

Electroporation is a complex, iterative, and nonlinear phenomenon that is often studied by numerical simulations. In recent years, tissue electroporation simulations have been performed using static models. However, the results of a static model simulation are restricted to a fixed protocol signature of the pulsed electric field. This paper describes a novel dynamic model of tissue electroporation that also includes tissue dispersion and temperature to allow time-domain simulations. We implemented the biological dispersion of potato tubers and thermal analysis in a commercial finite element method software. A cell electroporation model was adapted to account for the increase in tissue conductivity. The model yielded twelve parameters, divided into three dynamic states of electroporation. Thermal analysis describes the dependence of tissue conductivity on temperature. The model parameters were evaluated using experiments with vegetal tissue (Solanum tuberosum) under electrochemotherapy protocols. The proposed model can accurately predict the conductivity of tissue under electroporation from 10 kV/m to 100 kV/m. A negligible thermal effect was observed at 100 kV/m, with a 0.89 °C increase. We believe that the proposed model is suitable for describing the electroporation current on a tissue scale and also for providing a hint on the effects on the cell membrane.

Coronavirus disease (COVID-19) is an infectious disease discovered in 2019 and currently in outbreak across the world. Lung injury with severe respiratory failure is the leading cause of death in COVID-19, brought by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). However, there still lacks efficient treatment for COVID-19 induced lung injury and acute respiratory failure. Inhibition of Angiotensin-converting enzyme 2 (ACE2) caused by spike protein of SARS-CoV-2 is the most plausible mechanism of lung injury in COVID-19. We propose two candidate drugs, COL-3 (a chemically modified tetracycline) and CGP-60474 (a cyclin-dependent kinase inhibitor), for treating lung injuries in COVID-19, based on their abilities to reverse the gene expression patterns in HCC515 cells treated with ACE2 inhibitor and in human COVID-19 patient lung tissues. Further bioinformatics analysis shows that twelve significantly enriched pathways (P-value <0.05) overlap between HCC515 cells treated with ACE2 inhibitor and human COVID-19 patient lung tissues, including signaling pathways known to be associated with lung injury such as TNF signaling, MAPK signaling and Chemokine signaling pathways. All these twelve pathways are targeted in COL-3 treated HCC515 cells, in which genes such as RHOA, RAC2, FAS, CDC42 have reduced expression. CGP-60474 shares eleven of twelve pathways with COL-3 with common target genes such as RHOA. It also uniquely targets genes related to lung injury, such as CALR and MMP14. In summary, this study shows that ACE2 inhibition is likely part of the mechanisms leading to lung injury in COVID-19, and that compounds such as COL-3 and CGP-60474 have the potential as repurposed drugs for its treatment.

Congenital heart disease (CHD) encompasses a spectrum of cardiovascular structural abnormalities, often requiring customized treatment plans for individual patients. Computational modeling and analysis of these unique cardiac anatomies can improve diagnosis and treatment planning and may ultimately lead to improved outcomes. Deep learning (DL) methods have demonstrated the potential to enable efficient treatment planning by automating cardiac segmentation and mesh construction for patients with normal cardiac anatomies. However, CHDs are often rare, making it challenging to acquire sufficiently large patient cohorts for training such DL models. Generative modeling of cardiac anatomies has the potential to fill this gap via the generation of virtual cohorts; however, prior approaches were largely designed for normal anatomies and cannot readily capture the significant topological variations seen in CHD patients. Therefore, we propose a type- and shape-disentangled generative approach suitable to capture the wide spectrum of cardiac anatomies observed in different CHD types and synthesize differently shaped cardiac anatomies that preserve the unique topology for specific CHD types. Our DL approach represents generic whole heart anatomies with CHD type-specific abnormalities implicitly using signed distance fields (SDF) based on CHD type diagnosis, which conveniently captures divergent anatomical variations across different types and represents meaningful intermediate CHD states. To capture the shape-specific variations, we then learn invertible deformations to morph the learned CHD type-specific anatomies and reconstruct patient-specific shapes. Our approach has the potential to augment the image-segmentation pairs for rarer CHD types for cardiac segmentation and generate cohorts of CHD cardiac meshes for computational simulation.

Novel diagnostic and therapeutic radiopharmaceuticals are increasingly becoming a central part of personalized medicine. Continued innovation in the development of new radiopharmaceuticals is key to sustained growth and advancement of precision medicine. Artificial intelligence (AI) has been used in multiple fields of medicine to develop and validate better tools for patient diagnosis and therapy, including in radiopharmaceutical design. In this review, we first discuss common in silico approaches and focus on their utility and challenges in radiopharmaceutical development. Next, we discuss the practical applications of in silico modeling in design of radiopharmaceuticals in various diseases.

Large-scale tissue deformation which is fundamental to tissue development hinges on local cellular rearrangements, such as T1 transitions. In the realm of the multi-phase field model, we analyse the statistical and dynamical properties of T1 transitions in a confluent tissue. We identify an energy profile that is robust to changes in several model parameters. It is characterized by an asymmetric profile with a fast increase in energy before the T1 transition and a sudden drop after the T1 transition, followed by a slow relaxation. The latter being a signature of the fluidity of the cell tissue. We show that T1 transitions are sources of localised large deformation of the cells undergoing the neighbour exchange and induce other T1 transitions in the nearby cells through a chaining of events that propagate local cell deformation to large scale tissue flows.