Neurotransmitter classification from electron microscopy images at synaptic sites in Drosophila melanogaster.

High-resolution electron microscopy of nervous systems has enabled the reconstruction of synaptic connectomes. However, we do not know the synaptic sign for each connection (i.e., whether a connection is excitatory or inhibitory), which is implied by the released transmitter. We demonstrate that artificial neural networks can predict transmitter types for presynapses from electron micrographs: a network trained to predict six transmitters (acetylcholine, glutamate, GABA, serotonin, dopamine, octopamine) achieves an accuracy of 87% for individual synapses, 94% for neurons, and 91% for known cell types across a D. melanogaster whole brain. We visualize the ultrastructural features used for prediction, discovering subtle but significant differences between transmitter phenotypes. We also analyze transmitter distributions across the brain and find that neurons that develop together largely express only one fast-acting transmitter (acetylcholine, glutamate, or GABA). We hope that our publicly available predictions act as an accelerant for neuroscientific hypothesis generation for the fly.

Artificial Intelligence for Drug Discovery: Are We There Yet?

Drug discovery is adapting to novel technologies such as data science, informatics, and artificial intelligence (AI) to accelerate effective treatment development while reducing costs and animal experiments. AI is transforming drug discovery, as indicated by increasing interest from investors, industrial and academic scientists, and legislators. Successful drug discovery requires optimizing properties related to pharmacodynamics, pharmacokinetics, and clinical outcomes. This review discusses the use of AI in the three pillars of drug discovery: diseases, targets, and therapeutic modalities, with a focus on small-molecule drugs. AI technologies, such as generative chemistry, machine learning, and multiproperty optimization, have enabled several compounds to enter clinical trials. The scientific community must carefully vet known information to address the reproducibility crisis. The full potential of AI in drug discovery can only be realized with sufficient ground truth and appropriate human intervention at later pipeline stages.

Impossibility theorems for feature attribution.

Despite a sea of interpretability methods that can produce plausible explanations, the field has also empirically seen many failure cases of such methods. In light of these results, it remains unclear for practitioners how to use these methods and choose between them in a principled way. In this paper, we show that for moderately rich model classes (easily satisfied by neural networks), any feature attribution method that is complete and linear-for example, Integrated Gradients and Shapley Additive Explanations (SHAP)-can provably fail to improve on random guessing for inferring model behavior. Our results apply to common end-tasks such as characterizing local model behavior, identifying spurious features, and algorithmic recourse. One takeaway from our work is the importance of concretely defining end-tasks: Once such an end-task is defined, a simple and direct approach of repeated model evaluations can outperform many other complex feature attribution methods.

Deep learning models reveal replicable, generalizable, and behaviorally relevant sex differences in human functional brain organization.

Sex plays a crucial role in human brain development, aging, and the manifestation of psychiatric and neurological disorders. However, our understanding of sex differences in human functional brain organization and their behavioral consequences has been hindered by inconsistent findings and a lack of replication. Here, we address these challenges using a spatiotemporal deep neural network (stDNN) model to uncover latent functional brain dynamics that distinguish male and female brains. Our stDNN model accurately differentiated male and female brains, demonstrating consistently high cross-validation accuracy (>90%), replicability, and generalizability across multisession data from the same individuals and three independent cohorts (N ~ 1,500 young adults aged 20 to 35). Explainable AI (XAI) analysis revealed that brain features associated with the default mode network, striatum, and limbic network consistently exhibited significant sex differences (effect sizes > 1.5) across sessions and independent cohorts. Furthermore, XAI-derived brain features accurately predicted sex-specific cognitive profiles, a finding that was also independently replicated. Our results demonstrate that sex differences in functional brain dynamics are not only highly replicable and generalizable but also behaviorally relevant, challenging the notion of a continuum in male-female brain organization. Our findings underscore the crucial role of sex as a biological determinant in human brain organization, have significant implications for developing personalized sex-specific biomarkers in psychiatric and neurological disorders, and provide innovative AI-based computational tools for future research.

Interpreting chemisorption strength with AutoML-based feature deletion experiments.

The chemisorption energy of reactants on a catalyst surface, [Formula: see text], is among the most informative characteristics of understanding and pinpointing the optimal catalyst. The intrinsic complexity of catalyst surfaces and chemisorption reactions presents significant difficulties in identifying the pivotal physical quantities determining [Formula: see text]. In response to this, the study proposes a methodology, the feature deletion experiment, based on Automatic Machine Learning (AutoML) for knowledge extraction from a high-throughput density functional theory (DFT) database. The study reveals that, for binary alloy surfaces, the local adsorption site geometric information is the primary physical quantity determining [Formula: see text], compared to the electronic and physiochemical properties of the catalyst alloys. By integrating the feature deletion experiment with instance-wise variable selection (INVASE), a neural network-based explainable AI (XAI) tool, we established the best-performing feature set containing 21 intrinsic, non-DFT computed properties, achieving an MAE of 0.23 eV across a periodic table-wide chemical space involving more than 1,600 types of alloys surfaces and 8,400 chemisorption reactions. This study demonstrates the stability, consistency, and potential of AutoML-based feature deletion experiment in developing concise, predictive, and theoretically meaningful models for complex chemical problems with minimal human intervention.

xSiGra: explainable model for single-cell spatial data elucidation.

Recent advancements in spatial imaging technologies have revolutionized the acquisition of high-resolution multichannel images, gene expressions, and spatial locations at the single-cell level. Our study introduces xSiGra, an interpretable graph-based AI model, designed to elucidate interpretable features of identified spatial cell types, by harnessing multimodal features from spatial imaging technologies. By constructing a spatial cellular graph with immunohistology images and gene expression as node attributes, xSiGra employs hybrid graph transformer models to delineate spatial cell types. Additionally, xSiGra integrates a novel variant of gradient-weighted class activation mapping component to uncover interpretable features, including pivotal genes and cells for various cell types, thereby facilitating deeper biological insights from spatial data. Through rigorous benchmarking against existing methods, xSiGra demonstrates superior performance across diverse spatial imaging datasets. Application of xSiGra on a lung tumor slice unveils the importance score of cells, illustrating that cellular activity is not solely determined by itself but also impacted by neighboring cells. Moreover, leveraging the identified interpretable genes, xSiGra reveals endothelial cell subset interacting with tumor cells, indicating its heterogeneous underlying mechanisms within complex cellular interactions.

Deep learning to capture leaf shape in plant images: Validation by geometric morphometrics.

Plant leaves play a pivotal role in automated species identification using deep learning (DL). However, achieving reproducible capture of leaf variation remains challenging due to the inherent "black box" problem of DL models. To evaluate the effectiveness of DL in capturing leaf shape, we used geometric morphometrics (GM), an emerging component of eXplainable Artificial Intelligence (XAI) toolkits. We photographed Ranunculus auricomus leaves directly in situ and after herbarization. From these corresponding leaf images, we automatically extracted DL features using a neural network and digitized leaf shapes using GM. The association between the extracted DL features and GM shapes was then evaluated using dimension reduction and covariation models. DL features facilitated the clustering of leaf images by source populations in both in situ and herbarized leaf image datasets, and certain DL features were significantly associated with biological leaf shape variation as inferred by GM. DL features also enabled leaf classification into morpho-phylogenomic groups within the intricate R. auricomus species complex. We demonstrated that simple in situ leaf imaging and DL reproducibly captured leaf shape variation at the population level, while combining this approach with GM provided key insights into the shape information extracted from images by computer vision, a necessary prerequisite for reliable automated plant phenotyping.

Explainable AI for Bioinformatics: Methods, Tools and Applications.

Artificial intelligence (AI) systems utilizing deep neural networks and machine learning (ML) algorithms are widely used for solving critical problems in bioinformatics, biomedical informatics and precision medicine. However, complex ML models that are often perceived as opaque and black-box methods make it difficult to understand the reasoning behind their decisions. This lack of transparency can be a challenge for both end-users and decision-makers, as well as AI developers. In sensitive areas such as healthcare, explainability and accountability are not only desirable properties but also legally required for AI systems that can have a significant impact on human lives. Fairness is another growing concern, as algorithmic decisions should not show bias or discrimination towards certain groups or individuals based on sensitive attributes. Explainable AI (XAI) aims to overcome the opaqueness of black-box models and to provide transparency in how AI systems make decisions. Interpretable ML models can explain how they make predictions and identify factors that influence their outcomes. However, the majority of the state-of-the-art interpretable ML methods are domain-agnostic and have evolved from fields such as computer vision, automated reasoning or statistics, making direct application to bioinformatics problems challenging without customization and domain adaptation. In this paper, we discuss the importance of explainability and algorithmic transparency in the context of bioinformatics. We provide an overview of model-specific and model-agnostic interpretable ML methods and tools and outline their potential limitations. We discuss how existing interpretable ML methods can be customized and fit to bioinformatics research problems. Further, through case studies in bioimaging, cancer genomics and text mining, we demonstrate how XAI methods can improve transparency and decision fairness. Our review aims at providing valuable insights and serving as a starting point for researchers wanting to enhance explainability and decision transparency while solving bioinformatics problems. GitHub: https://github.com/rezacsedu/XAI-for-bioinformatics.

Explicable prioritization of genetic variants by integration of rule-based and machine learning algorithms for diagnosis of rare Mendelian disorders.

BACKGROUND: In the process of finding the causative variant of rare diseases, accurate assessment and prioritization of genetic variants is essential. Previous variant prioritization tools mainly depend on the in-silico prediction of the pathogenicity of variants, which results in low sensitivity and difficulty in interpreting the prioritization result. In this study, we propose an explainable algorithm for variant prioritization, named 3ASC, with higher sensitivity and ability to annotate evidence used for prioritization. 3ASC annotates each variant with the 28 criteria defined by the ACMG/AMP genome interpretation guidelines and features related to the clinical interpretation of the variants. The system can explain the result based on annotated evidence and feature contributions. RESULTS: We trained various machine learning algorithms using in-house patient data. The performance of variant ranking was assessed using the recall rate of identifying causative variants in the top-ranked variants. The best practice model was a random forest classifier that showed top 1 recall of 85.6% and top 3 recall of 94.4%. The 3ASC annotates the ACMG/AMP criteria for each genetic variant of a patient so that clinical geneticists can interpret the result as in the CAGI6 SickKids challenge. In the challenge, 3ASC identified causal genes for 10 out of 14 patient cases, with evidence of decreased gene expression for 6 cases. Among them, two genes (HDAC8 and CASK) had decreased gene expression profiles confirmed by transcriptome data. CONCLUSIONS: 3ASC can prioritize genetic variants with higher sensitivity compared to previous methods by integrating various features related to clinical interpretation, including features related to false positive risk such as quality control and disease inheritance pattern. The system allows interpretation of each variant based on the ACMG/AMP criteria and feature contribution assessed using explainable AI techniques.

One label is all you need: Interpretable AI-enhanced histopathology for oncology.

Artificial Intelligence (AI)-enhanced histopathology presents unprecedented opportunities to benefit oncology through interpretable methods that require only one overall label per hematoxylin and eosin (H&E) slide with no tissue-level annotations. We present a structured review of these methods organized by their degree of verifiability and by commonly recurring application areas in oncological characterization. First, we discuss morphological markers (tumor presence/absence, metastases, subtypes, grades) in which AI-identified regions of interest (ROIs) within whole slide images (WSIs) verifiably overlap with pathologist-identified ROIs. Second, we discuss molecular markers (gene expression, molecular subtyping) that are not verified via H&E but rather based on overlap with positive regions on adjacent tissue. Third, we discuss genetic markers (mutations, mutational burden, microsatellite instability, chromosomal instability) that current technologies cannot verify if AI methods spatially resolve specific genetic alterations. Fourth, we discuss the direct prediction of survival to which AI-identified histopathological features quantitatively correlate but are nonetheless not mechanistically verifiable. Finally, we discuss in detail several opportunities and challenges for these one-label-per-slide methods within oncology. Opportunities include reducing the cost of research and clinical care, reducing the workload of clinicians, personalized medicine, and unlocking the full potential of histopathology through new imaging-based biomarkers. Current challenges include explainability and interpretability, validation via adjacent tissue sections, reproducibility, data availability, computational needs, data requirements, domain adaptability, external validation, dataset imbalances, and finally commercialization and clinical potential. Ultimately, the relative ease and minimum upfront cost with which relevant data can be collected in addition to the plethora of available AI methods for outcome-driven analysis will surmount these current limitations and achieve the innumerable opportunities associated with AI-driven histopathology for the benefit of oncology.

Few-shot genes selection: subset of PAM50 genes for breast cancer subtypes classification.

BACKGROUND: In recent years, researchers have made significant strides in understanding the heterogeneity of breast cancer and its various subtypes. However, the wealth of genomic and proteomic data available today necessitates efficient frameworks, instruments, and computational tools for meaningful analysis. Despite its success as a prognostic tool, the PAM50 gene signature's reliance on many genes presents challenges in terms of cost and complexity. Consequently, there is a need for more efficient methods to classify breast cancer subtypes using a reduced gene set accurately. RESULTS: This study explores the potential of achieving precise breast cancer subtype categorization using a reduced gene set derived from the PAM50 gene signature. By employing a "Few-Shot Genes Selection" method, we randomly select smaller subsets from PAM50 and evaluate their performance using metrics and a linear model, specifically the Support Vector Machine (SVM) classifier. In addition, we aim to assess whether a more compact gene set can maintain performance while simplifying the classification process. Our findings demonstrate that certain reduced gene subsets can perform comparable or superior to the full PAM50 gene signature. CONCLUSIONS: The identified gene subsets, with 36 genes, have the potential to contribute to the development of more cost-effective and streamlined diagnostic tools in breast cancer research and clinical settings.

Slideflow: deep learning for digital histopathology with real-time whole-slide visualization.

Deep learning methods have emerged as powerful tools for analyzing histopathological images, but current methods are often specialized for specific domains and software environments, and few open-source options exist for deploying models in an interactive interface. Experimenting with different deep learning approaches typically requires switching software libraries and reprocessing data, reducing the feasibility and practicality of experimenting with new architectures. We developed a flexible deep learning library for histopathology called Slideflow, a package which supports a broad array of deep learning methods for digital pathology and includes a fast whole-slide interface for deploying trained models. Slideflow includes unique tools for whole-slide image data processing, efficient stain normalization and augmentation, weakly-supervised whole-slide classification, uncertainty quantification, feature generation, feature space analysis, and explainability. Whole-slide image processing is highly optimized, enabling whole-slide tile extraction at 40x magnification in 2.5 s per slide. The framework-agnostic data processing pipeline enables rapid experimentation with new methods built with either Tensorflow or PyTorch, and the graphical user interface supports real-time visualization of slides, predictions, heatmaps, and feature space characteristics on a variety of hardware devices, including ARM-based devices such as the Raspberry Pi.

Macroscopic brain dynamics beyond contralateral primary motor cortex for movement prediction.

This study investigates the complex relationship between upper limb movement direction and macroscopic neural signals in the brain, which is critical for understanding brain-computer interfaces (BCI). Conventional BCI research has primarily focused on a local area, such as the contralateral primary motor cortex (M1), relying on the population-based decoding method with microelectrode arrays. In contrast, macroscopic approaches such as electroencephalography (EEG) and magnetoencephalography (MEG) utilize numerous electrodes to cover broader brain regions. This study probes the potential differences in the mechanisms of microscopic and macroscopic methods. It is important to determine which neural activities effectively predict movements. To investigate this, we analyzed MEG data from nine right-handed participants while performing arm-reaching tasks. We employed dynamic statistical parametric mapping (dSPM) to estimate source activity and built a decoding model composed of long short-term memory (LSTM) and a multilayer perceptron to predict movement trajectories. This model achieved a high correlation coefficient of 0.79 between actual and predicted trajectories. Subsequently, we identified brain regions sensitive to predicting movement direction using the integrated gradients (IG) method, which assesses the predictive contribution of each source activity. The resulting salience map demonstrated a distribution without significant differences across motor-related regions, including M1. Predictions based solely on M1 activity yielded a correlation coefficient of 0.42, nearly half as effective as predictions incorporating all source activities. This suggests that upper limb movements are influenced by various factors such as movement coordination, planning, body and target position recognition, and control, beyond simple muscle activity. All of the activities are needed in the decoding model using macroscopic signals. Our findings also revealed that contralateral and ipsilateral hemispheres contribute equally to movement prediction, implying that BCIs could potentially benefit patients with brain damage in the contralateral hemisphere by utilizing brain signals from the ipsilateral hemisphere. In conclusion, this study demonstrates that macroscopic activity from large brain regions significantly contributes to predicting upper limb movement. Non-invasive BCI systems would require a comprehensive collection of neural signals from multiple brain regions.

Deep multimodal saliency parcellation of cerebellar pathways: Linking microstructure and individual function through explainable multitask learning.

Parcellation of human cerebellar pathways is essential for advancing our understanding of the human brain. Existing diffusion magnetic resonance imaging tractography parcellation methods have been successful in defining major cerebellar fibre tracts, while relying solely on fibre tract structure. However, each fibre tract may relay information related to multiple cognitive and motor functions of the cerebellum. Hence, it may be beneficial for parcellation to consider the potential importance of the fibre tracts for individual motor and cognitive functional performance measures. In this work, we propose a multimodal data-driven method for cerebellar pathway parcellation, which incorporates both measures of microstructure and connectivity, and measures of individual functional performance. Our method involves first training a multitask deep network to predict various cognitive and motor measures from a set of fibre tract structural features. The importance of each structural feature for predicting each functional measure is then computed, resulting in a set of structure-function saliency values that are clustered to parcellate cerebellar pathways. We refer to our method as Deep Multimodal Saliency Parcellation (DeepMSP), as it computes the saliency of structural measures for predicting cognitive and motor functional performance, with these saliencies being applied to the task of parcellation. Applying DeepMSP to a large-scale dataset from the Human Connectome Project Young Adult study (n = 1065), we found that it was feasible to identify multiple cerebellar pathway parcels with unique structure-function saliency patterns that were stable across training folds. We thoroughly experimented with all stages of the DeepMSP pipeline, including network selection, structure-function saliency representation, clustering algorithm, and cluster count. We found that a 1D convolutional neural network architecture and a transformer network architecture both performed comparably for the multitask prediction of endurance, strength, reading decoding, and vocabulary comprehension, with both architectures outperforming a fully connected network architecture. Quantitative experiments demonstrated that a proposed low-dimensional saliency representation with an explicit measure of motor versus cognitive category bias achieved the best parcellation results, while a parcel count of four was most successful according to standard cluster quality metrics. Our results suggested that motor and cognitive saliencies are distributed across the cerebellar white matter pathways. Inspection of the final k = 4 parcellation revealed that the highest-saliency parcel was most salient for the prediction of both motor and cognitive performance scores and included parts of the middle and superior cerebellar peduncles. Our proposed saliency-based parcellation framework, DeepMSP, enables multimodal, data-driven tractography parcellation. Through utilising both structural features and functional performance measures, this parcellation strategy may have the potential to enhance the study of structure-function relationships of the cerebellar pathways.

HistoEM: A Pathologist-Guided and Explainable Workflow Using Histogram Embedding for Gland Classification.

Pathologists have, over several decades, developed criteria for diagnosing and grading prostate cancer. However, this knowledge has not, so far, been included in the design of convolutional neural networks (CNN) for prostate cancer detection and grading. Further, it is not known whether the features learned by machine-learning algorithms coincide with diagnostic features used by pathologists. We propose a framework that enforces algorithms to learn the cellular and subcellular differences between benign and cancerous prostate glands in digital slides from hematoxylin and eosin-stained tissue sections. After accurate gland segmentation and exclusion of the stroma, the central component of the pipeline, named HistoEM, utilizes a histogram embedding of features from the latent space of the CNN encoder. Each gland is represented by 128 feature-wise histograms that provide the input into a second network for benign vs cancer classification of the whole gland. Cancer glands are further processed by a U-Net structured network to separate low-grade from high-grade cancer. Our model demonstrates similar performance compared with other state-of-the-art prostate cancer grading models with gland-level resolution. To understand the features learned by HistoEM, we first rank features based on the distance between benign and cancer histograms and visualize the tissue origins of the 2 most important features. A heatmap of pixel activation by each feature is generated using Grad-CAM and overlaid on nuclear segmentation outlines. We conclude that HistoEM, similar to pathologists, uses nuclear features for the detection of prostate cancer. Altogether, this novel approach can be broadly deployed to visualize computer-learned features in histopathology images.

UltraAIGenomics: Artificial Intelligence-Based Cardiovascular Disease Risk Assessment by Fusion of Ultrasound-Based Radiomics and Genomics Features for Preventive, Personalized and Precision Medicine: A Narrative Review.

Cardiovascular disease (CVD) diagnosis and treatment are challenging since symptoms appear late in the disease's progression. Despite clinical risk scores, cardiac event prediction is inadequate, and many at-risk patients are not adequately categorised by conventional risk factors alone. Integrating genomic-based biomarkers (GBBM), specifically those found in plasma and/or serum samples, along with novel non-invasive radiomic-based biomarkers (RBBM) such as plaque area and plaque burden can improve the overall specificity of CVD risk. This review proposes two hypotheses: (i) RBBM and GBBM biomarkers have a strong correlation and can be used to detect the severity of CVD and stroke precisely, and (ii) introduces a proposed artificial intelligence (AI)-based preventive, precision, and personalized ( aiP 3 ) CVD/Stroke risk model. The PRISMA search selected 246 studies for the CVD/Stroke risk. It showed that using the RBBM and GBBM biomarkers, deep learning (DL) modelscould be used for CVD/Stroke risk stratification in the aiP 3 framework. Furthermore, we present a concise overview of platelet function, complete blood count (CBC), and diagnostic methods. As part of the AI paradigm, we discuss explainability, pruning, bias, and benchmarking against previous studies and their potential impacts. The review proposes the integration of RBBM and GBBM, an innovative solution streamlined in the DL paradigm for predicting CVD/Stroke risk in the aiP 3 framework. The combination of RBBM and GBBM introduces a powerful CVD/Stroke risk assessment paradigm. aiP 3 model signifies a promising advancement in CVD/Stroke risk assessment.

A systematic review of machine learning models for management, prediction and classification of ARDS.

AIM: Acute respiratory distress syndrome or ARDS is an acute, severe form of respiratory failure characterised by poor oxygenation and bilateral pulmonary infiltrates. Advancements in signal processing and machine learning have led to promising solutions for classification, event detection and predictive models in the management of ARDS. METHOD: In this review, we provide systematic description of different studies in the application of Machine Learning (ML) and artificial intelligence for management, prediction, and classification of ARDS. We searched the following databases: Google Scholar, PubMed, and EBSCO from 2009 to 2023. A total of 243 studies was screened, in which, 52 studies were included for review and analysis. We integrated knowledge of previous work providing the state of art and overview of explainable decision models in machine learning and have identified areas for future research. RESULTS: Gradient boosting is the most common and successful method utilised in 12 (23.1%) of the studies. Due to limitation of data size available, neural network and its variation is used by only 8 (15.4%) studies. Whilst all studies used cross validating technique or separated database for validation, only 1 study validated the model with clinician input. Explainability methods were presented in 15 (28.8%) of studies with the most common method is feature importance which used 14 times. CONCLUSION: For databases of 5000 or fewer samples, extreme gradient boosting has the highest probability of success. A large, multi-region, multi centre database is required to reduce bias and take advantage of neural network method. A framework for validating with and explaining ML model to clinicians involved in the management of ARDS would be very helpful for development and deployment of the ML model.

AI-Enhanced Detection of Clinically Relevant Structural and Functional Anomalies in MRI: Traversing the Landscape of Conventional to Explainable Approaches.

Anomaly detection in medical imaging, particularly within the realm of magnetic resonance imaging (MRI), stands as a vital area of research with far-reaching implications across various medical fields. This review meticulously examines the integration of artificial intelligence (AI) in anomaly detection for MR images, spotlighting its transformative impact on medical diagnostics. We delve into the forefront of AI applications in MRI, exploring advanced machine learning (ML) and deep learning (DL) methodologies that are pivotal in enhancing the precision of diagnostic processes. The review provides a detailed analysis of preprocessing, feature extraction, classification, and segmentation techniques, alongside a comprehensive evaluation of commonly used metrics. Further, this paper explores the latest developments in ensemble methods and explainable AI, offering insights into future directions and potential breakthroughs. This review synthesizes current insights, offering a valuable guide for researchers, clinicians, and medical imaging experts. It highlights AI's crucial role in improving the precision and speed of detecting key structural and functional irregularities in MRI. Our exploration of innovative techniques and trends furthers MRI technology development, aiming to refine diagnostics, tailor treatments, and elevate patient care outcomes. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY: Stage 1.

Performance of explainable artificial intelligence in guiding the management of patients with a pancreatic cyst.

BACKGROUND/OBJECTIVES: Pancreatic cyst management can be distilled into three separate pathways - discharge, monitoring or surgery- based on the risk of malignant transformation. This study compares the performance of artificial intelligence (AI) models to clinical care for this task. METHODS: Two explainable boosting machine (EBM) models were developed and evaluated using clinical features only, or clinical features and cyst fluid molecular markers (CFMM) using a publicly available dataset, consisting of 850 cases (median age 64; 65 % female) with independent training (429 cases) and holdout test cohorts (421 cases). There were 137 cysts with no malignant potential, 114 malignant cysts, and 599 IPMNs and MCNs. RESULTS: The EBM and EBM with CFMM models had higher accuracy for identifying patients requiring monitoring (0.88 and 0.82) and surgery (0.66 and 0.82) respectively compared with current clinical care (0.62 and 0.58). For discharge, the EBM with CFMM model had a higher accuracy (0.91) than either the EBM model (0.84) or current clinical care (0.86). In the cohort of patients who underwent surgical resection, use of the EBM-CFMM model would have decreased the number of unnecessary surgeries by 59 % (n = 92), increased correct surgeries by 7.5 % (n = 11), identified patients who require monitoring by 122 % (n = 76), and increased the number of patients correctly classified for discharge by 138 % (n = 18) compared to clinical care. CONCLUSIONS: EBM models had greater sensitivity and specificity for identifying the correct management compared with either clinical management or previous AI models. The model predictions are demonstrated to be interpretable by clinicians.

CLARUS: An interactive explainable AI platform for manual counterfactuals in graph neural networks.

BACKGROUND: Lack of trust in artificial intelligence (AI) models in medicine is still the key blockage for the use of AI in clinical decision support systems (CDSS). Although AI models are already performing excellently in systems medicine, their black-box nature entails that patient-specific decisions are incomprehensible for the physician. Explainable AI (XAI) algorithms aim to "explain" to a human domain expert, which input features influenced a specific recommendation. However, in the clinical domain, these explanations must lead to some degree of causal understanding by a clinician. RESULTS: We developed the CLARUS platform, aiming to promote human understanding of graph neural network (GNN) predictions. CLARUS enables the visualisation of patient-specific networks, as well as, relevance values for genes and interactions, computed by XAI methods, such as GNNExplainer. This enables domain experts to gain deeper insights into the network and more importantly, the expert can interactively alter the patient-specific network based on the acquired understanding and initiate re-prediction or retraining. This interactivity allows us to ask manual counterfactual questions and analyse the effects on the GNN prediction. CONCLUSION: We present the first interactive XAI platform prototype, CLARUS, that allows not only the evaluation of specific human counterfactual questions based on user-defined alterations of patient networks and a re-prediction of the clinical outcome but also a retraining of the entire GNN after changing the underlying graph structures. The platform is currently hosted by the GWDG on https://rshiny.gwdg.de/apps/clarus/.