Explainable artificial intelligence (XAI) elucidates the decision-making process of complex AI models and is important in building trust in model predictions. XAI explanations themselves require evaluation as to accuracy and reasonableness and in the context of use of the underlying AI model. This review details the evaluation of XAI in cardiac AI applications and has found that, of the studies examined, 37% evaluated XAI quality using literature results, 11% used clinicians as domain-experts, 11% used proxies or statistical analysis, with the remaining 43% not assessing the XAI used at all. We aim to inspire additional studies within healthcare, urging researchers not only to apply XAI methods but to systematically assess the resulting explanations, as a step towards developing trustworthy and safe models.
UNASSIGNED: The online version contains supplementary material available at 10.1007/s10462-024-10852-w.

Metastatic breast cancer (MBC) continues to be a leading cause of cancer-related deaths among women. This work introduces an innovative non-invasive breast cancer classification model designed to improve the identification of cancer metastases. While this study marks the initial exploration into predicting MBC, additional investigations are essential to validate the occurrence of MBC. Our approach combines the strengths of large language models (LLMs), specifically the bidirectional encoder representations from transformers (BERT) model, with the powerful capabilities of graph neural networks (GNNs) to predict MBC patients based on their histopathology reports. This paper introduces a BERT-GNN approach for metastatic breast cancer prediction (BG-MBC) that integrates graph information derived from the BERT model. In this model, nodes are constructed from patient medical records, while BERT embeddings are employed to vectorise representations of the words in histopathology reports, thereby capturing semantic information crucial for classification by employing three distinct approaches (namely univariate selection, extra trees classifier for feature importance, and Shapley values to identify the features that have the most significant impact). Identifying the most crucial 30 features out of 676 generated as embeddings during model training, our model further enhances its predictive capabilities. The BG-MBC model achieves outstanding accuracy, with a detection rate of 0.98 and an area under curve (AUC) of 0.98, in identifying MBC patients. This remarkable performance is credited to the model\'s utilisation of attention scores generated by the LLM from histopathology reports, effectively capturing pertinent features for classification.

With the outbreak of COVID-19 in 2020, countries worldwide faced significant concerns and challenges. Various studies have emerged utilizing Artificial Intelligence (AI) and Data Science techniques for disease detection. Although COVID-19 cases have declined, there are still cases and deaths around the world. Therefore, early detection of COVID-19 before the onset of symptoms has become crucial in reducing its extensive impact. Fortunately, wearable devices such as smartwatches have proven to be valuable sources of physiological data, including Heart Rate (HR) and sleep quality, enabling the detection of inflammatory diseases. In this study, we utilize an already-existing dataset that includes individual step counts and heart rate data to predict the probability of COVID-19 infection before the onset of symptoms. We train three main model architectures: the Gradient Boosting classifier (GB), CatBoost trees, and TabNet classifier to analyze the physiological data and compare their respective performances. We also add an interpretability layer to our best-performing model, which clarifies prediction results and allows a detailed assessment of effectiveness. Moreover, we created a private dataset by gathering physiological data from Fitbit devices to guarantee reliability and avoid bias.The identical set of models was then applied to this private dataset using the same pre-trained models, and the results were documented. Using the CatBoost tree-based method, our best-performing model outperformed previous studies with an accuracy rate of 85% on the publicly available dataset. Furthermore, this identical pre-trained CatBoost model produced an accuracy of 81% when applied to the private dataset. You will find the source code in the link: https://github.com/OpenUAE-LAB/Covid-19-detection-using-Wearable-data.git .

BACKGROUND: Responsible artificial intelligence (RAI) emphasizes the use of ethical frameworks implementing accountability, responsibility, and transparency to address concerns in the deployment and use of artificial intelligence (AI) technologies, including privacy, autonomy, self-determination, bias, and transparency. Standards are under development to guide the support and implementation of AI given these considerations.
OBJECTIVE: The purpose of this review is to provide an overview of current research evidence and knowledge gaps regarding the implementation of RAI principles and the occurrence and resolution of ethical issues within AI systems.
METHODS: A scoping review following Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) guidelines was proposed. PubMed, ERIC, Scopus, IEEE Xplore, EBSCO, Web of Science, ACM Digital Library, and ProQuest (Arts and Humanities) will be systematically searched for articles published since 2013 that examine RAI principles and ethical concerns within AI. Eligibility assessment will be conducted independently and coded data will be analyzed along themes and stratified across discipline-specific literature.
RESULTS: The results will be included in the full scoping review, which is expected to start in June 2024 and completed for the submission of publication by the end of 2024.
CONCLUSIONS: This scoping review will summarize the state of evidence and provide an overview of its impact, as well as strengths, weaknesses, and gaps in research implementing RAI principles. The review may also reveal discipline-specific concerns, priorities, and proposed solutions to the concerns. It will thereby identify priority areas that should be the focus of future regulatory options available, connecting theoretical aspects of ethical requirements for principles with practical solutions.
UNASSIGNED: PRR1-10.2196/52349.

The local interpretable model-agnostic explanation (LIME) method was used to interpret two machine learning models of compounds penetrating the blood-brain barrier. The classification models, Random Forest, ExtraTrees, and Deep Residual Network, were trained and validated using the blood-brain barrier penetration dataset, which shows the penetrability of compounds in the blood-brain barrier. LIME was able to create explanations for such penetrability, highlighting the most important substructures of molecules that affect drug penetration in the barrier. The simple and intuitive outputs prove the applicability of this explainable model to interpreting the permeability of compounds across the blood-brain barrier in terms of molecular features. LIME explanations were filtered with a weight equal to or greater than 0.1 to obtain only the most relevant explanations. The results showed several structures that are important for blood-brain barrier penetration. In general, it was found that some compounds with nitrogenous substructures are more likely to permeate the blood-brain barrier. The application of these structural explanations may help the pharmaceutical industry and potential drug synthesis research groups to synthesize active molecules more rationally.

Marker-assisted selection (MAS) plays a crucial role in crop breeding improving the speed and precision of conventional breeding programmes by quickly and reliably identifying and selecting plants with desired traits. However, the efficacy of MAS depends on several prerequisites, with precise phenotyping being a key aspect of any plant breeding programme. Recent advancements in high-throughput remote phenotyping, facilitated by unmanned aerial vehicles coupled to machine learning, offer a non-destructive and efficient alternative to traditional, time-consuming, and labour-intensive methods. Furthermore, MAS relies on knowledge of marker-trait associations, commonly obtained through genome-wide association studies (GWAS), to understand complex traits such as drought tolerance, including yield components and phenology. However, GWAS has limitations that artificial intelligence (AI) has been shown to partially overcome. Additionally, AI and its explainable variants, which ensure transparency and interpretability, are increasingly being used as recognised problem-solving tools throughout the breeding process. Given these rapid technological advancements, this review provides an overview of state-of-the-art methods and processes underlying each MAS, from phenotyping, genotyping and association analyses to the integration of explainable AI along the entire workflow. In this context, we specifically address the challenges and importance of breeding winter wheat for greater drought tolerance with stable yields, as regional droughts during critical developmental stages pose a threat to winter wheat production. Finally, we explore the transition from scientific progress to practical implementation and discuss ways to bridge the gap between cutting-edge developments and breeders, expediting MAS-based winter wheat breeding for drought tolerance.

Seizure detection using machine learning is a critical problem for the timely intervention and management of epilepsy. We propose SeizFt, a robust seizure detection framework using EEG from a wearable device. It uses features paired with an ensemble of trees, thus enabling further interpretation of the model\'s results. The efficacy of the underlying augmentation and class-balancing strategy is also demonstrated. This study was performed for the Seizure Detection Challenge 2023, an ICASSP Grand Challenge.

Cancer, a life-threatening disorder caused by genetic abnormalities and metabolic irregularities, is a substantial health danger, with lung and colon cancer being major contributors to death. Histopathological identification is critical in directing effective treatment regimens for these cancers. The earlier these disorders are identified, the lesser the risk of death. The use of machine learning and deep learning approaches has the potential to speed up cancer diagnosis processes by allowing researchers to analyse large patient databases quickly and affordably. This study introduces the Inception-ResNetV2 model with strategically incorporated local binary patterns (LBP) features to improve diagnostic accuracy for lung and colon cancer identification. The model is trained on histopathological images, and the integration of deep learning and texture-based features has demonstrated its exceptional performance with 99.98% accuracy. Importantly, the study employs explainable artificial intelligence (AI) through SHapley Additive exPlanations (SHAP) to unravel the complex inner workings of deep learning models, providing transparency in decision-making processes. This study highlights the potential to revolutionize cancer diagnosis in an era of more accurate and reliable medical assessments.

Stakeholders of machine learning models desire explainable artificial intelligence (XAI) to produce human-understandable and consistent interpretations. In computational toxicity, augmentation of text-based molecular representations has been used successfully for transfer learning on downstream tasks. Augmentations of molecular representations can also be used at inference to compare differences between multiple representations of the same ground-truth. In this study, we investigate the robustness of eight XAI methods using test-time augmentation for a molecular-representation model in the field of computational toxicity prediction. We report significant differences between explanations for different representations of the same ground-truth, and show that randomized models have similar variance. We hypothesize that text-based molecular representations in this and past research reflect tokenization more than learned parameters. Furthermore, we see a greater variance between in-domain predictions than out-of-domain predictions, indicating XAI measures something other than learned parameters. Finally, we investigate the relative importance given to expert-derived structural alerts and find similar importance given irregardless of applicability domain, randomization and varying training procedures. We therefore caution future research to validate their methods using a similar comparison to human intuition without further investigation. SCIENTIFIC CONTRIBUTION: In this research we critically investigate XAI through test-time augmentation, contrasting previous assumptions about using expert validation and showing inconsistencies within models for identical representations. SMILES augmentation has been used to increase model accuracy, but was here adapted from the field of image test-time augmentation to be used as an independent indication of the consistency within SMILES-based molecular representation models.

Convolutional Neural Networks (CNNs) are frequently and successfully used in medical prediction tasks. They are often used in combination with transfer learning, leading to improved performance when training data for the task are scarce. The resulting models are highly complex and typically do not provide any insight into their predictive mechanisms, motivating the field of \"explainable\" artificial intelligence (XAI). However, previous studies have rarely quantitatively evaluated the \"explanation performance\" of XAI methods against ground-truth data, and transfer learning and its influence on objective measures of explanation performance has not been investigated. Here, we propose a benchmark dataset that allows for quantifying explanation performance in a realistic magnetic resonance imaging (MRI) classification task. We employ this benchmark to understand the influence of transfer learning on the quality of explanations. Experimental results show that popular XAI methods applied to the same underlying model differ vastly in performance, even when considering only correctly classified examples. We further observe that explanation performance strongly depends on the task used for pre-training and the number of CNN layers pre-trained. These results hold after correcting for a substantial correlation between explanation and classification performance.