Case-based reasoning in healthcare: Applications and challenges

Background: Breast cancer is one of the top causes of death for women worldwide with an incidence rate of about one in eight women. Mammographic screening allows earlier detection and offers the best chance of cure. The advances in artificial intelligence through Explainable AI, Machine Learning and Deep Learning have bettered diagnostic capacities but mammography remains the primary method for screening. Objectives: To introduce a structured comparative framework mapping XAI methods against datasets, model architectures, and reported performance which is absent in prior studies. Furthermore, the study incorporates emerging paradigms such as curriculum learning, federated learning, and case-based reasoning in mammography interpretation, which have not been jointly reviewed before. Method: This systematic scoping review examines the use of XAI in breast cancer identification and risk estimation. We performed systematic searches on Scopus, IEEE Explore, PubMed, Springer nature and Google Scholar following a systematic search strategy. The search was conducted through 2019 to early 2025, in peer‐reviewed studies of XAI methods in datasets of breast cancer. Findings: The implementation of Federated learning with hybrid ML methods enhances accuracy while preserving data privacy, while SHAP and LIME provide feature-based interpretability. Deep learning models like XGBoost and CNNs achieve high classification performance, and Grad-CAM aids radiologists by visually localizing tumours for better decision-making. Novelty: This review offers a comparative, metric-driven synthesis of XAI methods in mammography. It summarizes the comparison of datasets, model performance, and interpretability of the models. It also introduces emerging trends such as federated learning, curriculum learning, and human-centric AI, providing new perspectives on connecting algorithmic breakthroughs to clinical needs. Significance: XAI enhances the transparency, interpretability, fairness, and trustworthiness of the AI‐enabled health system and medical devices, and more importantly, quality of care and health outcomes. Keywords: Artificial Intelligence, Breast Cancer, Deep Learning, Explainable Artificial Intelligence (XAI), Grad-CAM, Interpretability, LIME, Machine Learning, SHAP, XGBoost

Legal documents are known to be long and complicated, which makes it essentially impossible for legal practitioners and researchers to quickly identify and extract relevant information. Here, a hybrid approach is presented that outperforms prior extractive and abstractive baselines on both lexical overlap and domain-specific reasoning metrics, which uses Case-Based Reasoning (CBR) for legal texts and concrete deep learning techniques for summary representation, accurately and efficiently producing summaries. Using a larger dataset of 4,968 legal cases from Kaggle, a multi-stage transformer architecture was constructed on top of the general CBR retrieval model created before to produce brief summaries along with CBR for context comprehension. The system was evaluated on legal outcome prediction and coherence of summary, with results showing performance superior to existing extractive and abstractive methods and trained the proposed model until 98% accuracy of legal entities, along with 46% more coherent legal corpus (baseline-enhanced) than state-of-the-art methods, compared using ROUGE scores above previous types by 23%. This study presents a hybrid legal text summarization framework that integrates CBR with transformer-based models. Extensive experiments show superior performance over recent baselines, achieving higher factual accuracy, reasoning fidelity, and legal entity preservation.

This systematic review synthesizes empirical evidence on artificial intelligence-enhanced learning interventions targeting executive function development across diverse populations and developmental stages within digital humanities contexts. Following PRISMA guidelines, a comprehensive search of five databases (PsycINFO, ERIC, Web of Science, Scopus, PubMed) from January 2020 through December 2024 identified 14 studies encompassing 1,810 participants aged 6 to 77 years. Included studies examined adaptive intelligent tutoring systems, virtual reality platforms, computerized cognitive training programs, computational thinking interventions, and machine learning-based assessment tools applied to humanities education and research. Results demonstrated consistent positive effects on inhibitory control (effect sizes: 0.11–0.62), cognitive flexibility, working memory (effect sizes: 0.09–0.18), and planning abilities, with machine learning models achieving high diagnostic accuracy (86.8%) for executive function impairments. Effectiveness was moderated by individual baseline cognitive capacity, particularly working memory constraints. Theoretical mechanisms underlying improvements included adaptive difficulty adjustment, cognitive load optimization, personalized scaffolding through Case-Based Reasoning and reinforcement learning algorithms, and neuroplasticity-driven efficiency gains. Despite promising findings, limitations include intervention heterogeneity, brief intervention durations, and limited long-term follow-up. Future research should prioritize longitudinal randomized controlled trials, neuroimaging studies elucidating neural mechanisms, and implementation science investigations supporting evidence-based integration of AI technologies in digital humanities pedagogy and clinical contexts.

This framework, “BQ-CBRS,” Hybrid Bigger Query-Case Based Reasoning System model, is the first of its kind to unite contextual embedding-based query expansion (using BERT), IndRNN-based semantic similarity scoring, the fuzzification of uncertain parameters, and XGBoost classification within one application to support precision agriculture. Some of the steps include query preprocessing, generating contextual embeddings utilizing a pre-trained method (for example, BERT), semantic similarity scoring using IndRNN, and expanding the query by adding top-ranked search terms. Fuzzification will acknowledge any uncertainty present in the data, while XGBoost will enhance the predictive power and efficacy of the present work. The proposed methodology consists of query preprocessing, contextual representations using pre-trained models (like BERT), calculating a similarity score through IndRNN, and expanding the query according to the top-scoring terms. Fuzzification will address the uncertainty in the data, and XGBoost will enhance prediction accuracy and efficiency. The Crop Recommendation Dataset consists of parameters, such as nitrogen, phosphorus, pH, temperature, and rainfall. The present model has low accuracy and low mean square error (MSE). Also, it improves over traditional approaches. The model will utilize precision agriculture technology to link historical cases and improve approaches for more effective resource management and advancing sustainable farming. This combination of symbolic reasoning and deep learning in the agriculture domain is novel, establishing a generalizable framework for intelligent decision support in dynamic and uncertain situations.

BackgroundAntimicrobial resistance (AMR) is a leading global health challenge responsible for over 1.27 million deaths annually and rising clinical and economic burdens on healthcare systems. Inappropriate prescribing is the primary driver of resistance, and preparing future healthcare professionals to prescribe responsibly is important. Antimicrobial stewardship (AMS) provides a structured approach to optimizing antimicrobial use and is recognized by the WHO as a priority. Undergraduate training offers a critical opportunity to embed stewardship principles before graduates assume independent roles. Despite this, AMS education remains inconsistent, often limited to factual teaching, with variable coverage across medicine, pharmacy, nursing, dentistry, veterinary medicine and midwifery. Applying a behavioural science framework such as COM-B enables evaluation of whether interventions target the capability, opportunity and motivation influences that underpin stewardship practice.ObjectivesThis review aimed to (i) identify AMS educational interventions delivered to undergraduate and pre-registration healthcare students across disciplines, and (ii) evaluate the extent to which they addressed behavioural coverage across COM-B influences necessary to prepare students for stewardship practice.MethodsA protocol was registered on PROSPERO (CRD420250655653). Six databases (PubMed, Web of Science, PsycINFO, EMBASE, CINAHL Plus, Scopus) were searched on 4 February 2025. Peer-reviewed studies in English evaluating AMS education for pre-registration students in medicine, pharmacy, nursing, dentistry, veterinary medicine and midwifery were included. Eligible interventions occurred in university or supervised clinical settings and reported at least one learning outcome relevant to stewardship. Data extraction was conducted in duplicate, with quality assessed using an adapted Medical Education Research Study Quality Instrument (MERSQI). A narrative synthesis was undertaken, coding interventions against COM-B domains adapted to an educational context.ResultsOf 7771 records screened, 42 studies were included, involving 8567 students across six continents. Mot used pre/post designs; two were randomized controlled trials. All addressed psychological capability (42/42), typically through lectures, online modules, or case-based reasoning tasks. Reflective motivation was supported in 25/42, often by strengthening prescribing confidence or professional identity. Fewer interventions created physical opportunity (20/42), such as access to prescribing charts, susceptibility data, or digital tools, or social opportunity (18/42) through teamwork and supervision. Only 9/42 developed physical capability via supervised practice or simulation. Just 2/42 addressed automatic motivation, such as the stress of time-pressured escape rooms or the satisfaction from user-friendly prescribing apps, showing how emotions can shape whether and how stewardship behaviours are carried out. Overall methodological quality was moderate (MERSQI mean=10/18).ConclusionsUndergraduate AMS education is widespread but uneven in its behavioural coverage across COM-B influences. Interventions predominantly focus on knowledge, with less attention to the skills, opportunities and motivational influences that enable stewardship in practice. COM-B highlights the need for curricula that combine factual teaching with rehearsal of practical skills, authentic resources, collaborative teamwork, reflective role identity and positive engagement. Embedding behavioural science principles into AMS education will better equip graduates to meet the complex stewardship demands of future healthcare.

Product design methods based on case-based reasoning (CBR) generally lack sufficient consideration of customer opinions. To address this problem, a user satisfaction-oriented CBR method for product conceptual design is proposed. The method consists of three stages. First, a quantitative mapping relationship between design cases and user satisfaction is established, and a case base is constructed. Subsequently, new product structural solutions are generated by adopting case reuse or innovative design approaches. Finally, a multi-objective decision-making model incorporating multi-dimensional factors is established to solve for the optimal conceptual design scheme. A conceptual design project for a tablet computer demonstrates that the proposed method effectively enhances user satisfaction with design schemes. This method provides a new perspective for the research of case-based product design methods and is of great value for improving the design symmetry between the conceptual design scheme and user requirements, and thereby enhancing the quality of enterprise product conceptual design.

ABSTRACT Traditional data‐driven optimization methods using case‐based reasoning (CBR) rely on heuristic similarity matching and lack probabilistic rigor, especially in complex processes like fluid catalytic cracking (FCC) with high dimensionality and uncertainty. To address these challenges, a novel data‐driven framework that integrates compact posterior estimation with CBR is proposed. The method first identifies key variables affecting product yields through information‐theoretic dimensionality reduction. Optimal operating parameters are then inferred using a combination of K‐nearest neighbors for similarity matching and Markov Chain Monte Carlo sampling for probabilistic estimation. Industrial validation showed gasoline and total liquid yields increased by 7.31% and 6.94%, respectively, with coke yield reduced by 5.83%. This approach successfully improves computational efficiency and optimization accuracy in practical applications.

Improving prototypical parts abstraction for case-based reasoning explanations designed for the kidney stone type recognition.

Combining case-based reasoning and process mining for root cause analysis in complex manufacturing environments

CBR-FoX: A case-based reasoning software tool for auditing time series predictions

Early and accurate triage of traumatic brain injury is critical for guiding treatment decisions that optimize patient outcomes. A major early clinical decision point occurs in the emergency department, where providers must decide whether to admit or discharge patients with head injuries, yet these decisions are often inconsistent and rarely supported by case-based frameworks. Here, we introduce RAPID-TBI (Retrieval Augmented Prediction for Informed Disposition in Traumatic Brain Injury), a multimodal system that predicts emergency department disposition using example-based retrieval to emulate clinical case-based reasoning. RAPID-TBI achieves state-of-the-art classification performance while enhancing interpretability by retrieving similar patients to inform predictions. Using a large multimodal TBI dataset from a major U.S. hospital system, RAPID-TBI integrates head CT scans, radiology reports, exam findings, laboratory values, vitals, and demographics through an attention-based encoder that generates patient embeddings for disposition classification. We further assessed RAPID-TBI across institutional and temporal generalizability, showing consistent performance and resilience to shifts in data distribution. Finally, we explored small language models as prompt-based classifiers for retrieval-guided prediction without fine-tuning. Together, these components enable RAPID-TBI to deliver consistent, individualized, and clinically grounded predictions, a promising step toward trustworthy, personalized decision support in TBI care.

Potatoes is an agricultural commodity that serves as a substitute for staple food. All kinds of ways have been done to increase the productivity of potato plants, but obstacles encountered in the process of planting potatoes include the presence of diseases that often result in crop failure. Lack of knowledge of farmers and the public about the types of diseases contained in potato plants, resulting in crop failure. Accurate and timely diagnosis of these diseases is essential for effective management and control. This paper presents the development of an expert system to diagnose diseases in potato plants using the Case Based Reasoning (CBR) method combined with the K-Nearest Neighbor (K-NN) algorithm. The system utilizes a database of past cases to identify and diagnose diseases based on the similarity between new cases and existing cases. The integration of CBR with K-NN algorithm improves the accuracy and reliability of diagnosis by considering various symptom features and environmental conditions. The results show that the system achieves a high level of accuracy in diagnosing potato plant diseases, outperforming traditional methods. This research aims to develop an efficient and easy-to-use tool for farmers and agricultural professionals to facilitate early detection and management of potato crop diseases, while improving the system performance metrics, including accuracy, precision, recall, and F1-Score, to assess the effectiveness of the system diagnostics for potato crop diseases. The contribution of this research aims to offer an easy accessible option for farmers to quickly identify and manage diseases in potato plants, thereby reducing losses due to crop failure. A future work will focus on expanding the system database and incorporating additional Machine Learning (ML) techniques to further improve diagnostic capabilities.

Reliable intrusion detection is critical for modern infrastructure security, yet it faces two fundamental challenges: scarcity of labeled samples and lack of model interpretability. Distributed optical fiber vibration sensing (DVS) systems are promising for perimeter security but perform poorly when only a few intrusion samples are available. Most deep models also lack transparency and trustworthiness. To address these issues, we propose an explainable dual-branch feature fusion dynamic class center prototypical network (DBFF-DC-ProtoNet). The framework employs a lightweight dual-branch 1-D ResNet to extract complementary temporal and time-frequency representations from raw signals and discrete wavelet transform (DWT) features, which are fused to form more discriminative class prototypes. A dynamic class center update strategy with a novel loss function is further introduced to enhance intra-class compactness and inter-class separability in few-shot conditions. In addition, an explainability module integrates prototype-based class activation mapping (Proto-CAM) and case-based reasoning, offering both fine-grained attribution of key signal segments and intuitive retrieval of similar historical cases. Extensive experiments on a self-collected dataset and a public benchmark confirm the effectiveness of our approach, achieving 97.22% and 98.33% accuracy under the 5-shot setting. These results demonstrate that DBFF-DC-ProtoNet effectively bridges few-shot learning with interpretability, providing a practical and trustworthy solution for DVS-based intrusion detection.

A hybrid electric motor equipment health management method based on case-based reasoning and fuzzy association rule mining

Abstract Earthquakes frequently occur worldwide and cause extensive damage. Effective emergency response is crucial, and accurate forecasting of emergency supply needs is essential for efficient disaster relief operations. This study proposes an emergency supply demand prediction model for high-altitude regions by combining case-based reasoning (CBR) with a radial basis function (RBF) neural network. Building on previous research into factors influencing earthquake casualties, we identified six key parameters, namely, the magnitude, seismic intensity, seismic fortification intensity, affected population, population density, and the number of damaged houses. After evaluating various machine learning models, the RBF neural network was found to have the highest accuracy in predicting earthquake casualties. Then, the model was integrated with CBR to further estimate emergency supply demand, with regional adjustments made according to the geological structures and specific supply needs of the western Sichuan plateau. To enhance practical usability, a graphical user interface was developed, enabling users to input earthquake parameters and obtain predictions of casualties and supply requirements. This research provides a valuable decision-support tool for earthquake emergency response in high-altitude regions and offers methodological insights for emergency supply demand prediction in other complex environments.

While machine learning (ML) techniques have been applied to detection and prediction tasks in clinical data, most methods rely on high-resolution data, which is not routinely available in most Intensive Care Units (ICUs), and perform poorly when faced with class imbalance. Here, we introduce and validate Collaborative Learning and Inference (CLaI) for detection and prediction of events from learned latent representations of multivariate physiological time series, leveraging similarities across patients. Our method offers a new way to detect and predict events using low-resolution physiological time series. We evaluate its performance on predicting intracranial hypertension and sepsis using the KidsBrainIT (minute-by-minute resolution) and MIMIC-IV (hourly resolution) datasets, respectively, comparing our approach with classification-based and sequence-to-sequence benchmarks from existing studies. Additional experiments on sepsis detection, robustness to class imbalance, and generalizability-demonstrated via seizure detection using the CHB-MIT scalp electroencephalogram dataset-confirm that CLaI effectively handles class imbalance, consistently achieving competitive performance and the highest F1 score. Overall, our approach introduces a novel method for analyzing routinely collected ICU physiological time series by leveraging patient similarity thus enabling ML interpretability through case-based reasoning.

A construction cost estimation system based on case-based reasoning and exponential smoothing

The object of research is the processes of case-based management a set of interconnected end-to-end business processes of the enterprise. The study is devoted to solving the problem of case-based management of interconnected end-to-end business processes of the enterprise that use shared resources. Research in this area is aimed at developing models, methods and technologies used in the management of business processes of the enterprise. The goal and main limitations of functional and process management in the form of a set of business processes that integrate the activities of the relevant divisions of the enterprise are determined and formally described. The main disadvantage of such management is associated with the mismatch between the existing organizational structure of the enterprise and end-to-end business processes that cover several of its divisions. Therefore, a transition from process to end-to-end business process management that use shared resources is proposed. This approach involves searching for and adapting of case-based, applying it and further preserving it. In conditions of restrictions on the execution of business processes, the use of a case-based reasoning allows increasing the efficiency of process management. An improvement of the method of case-based management of a group of end-to-end business processes is proposed. Unlike the existing one, it allows to determine the priorities of their access to resources, taking into account the restrictions on the time of their execution. This ensures the execution of processes within the established deadlines, which improves the economic performance of the enterprise. Practical application of the proposed improved method of case-based management of a group of end-to-end business processes allows to adjust the sequences of orders launch orders. This is done taking into account the restrictions on the execution time of each of the business processes, which allows to improve the process of order management at the enterprise.

Purpose Effective response to safety risks is critical to reducing the incidence of safety accidents and is an important element of project management on construction sites. The objective of this study is to propose a hybrid reasoning framework for generating safety risk response strategies based on construction safety requirements (SRs) and incorporating specific measures for effective response to potential risks. Design/methodology/approach This study proposes a hybrid reasoning framework based on SRs for generating risk response strategies that align with safety risk characteristics. The framework integrates rule-based reasoning (RBR) and case-based reasoning (CBR) while leveraging knowledge transformation and the reuse of domain ontologies to enhance the effectiveness of safety risk responses. Additionally, natural language processing (NLP) techniques are employed to identify safety risk factors, which are then analyzed using the multidimensional attribute Apriori (MA-Apriori) algorithm to determine their association with SRs. Meanwhile, RBR is used to design three types of reasoning rules by combining the relationship between safety risks and different classes in the domain ontology, and then case similarity is measured by CBR to obtain the risk response strategies under the corresponding reasoning rules. Findings A total of 209 accident case files were used for the testing of safety risk responses, and 34 safety risk factors and corresponding reasoning rules were obtained from the source cases with high relevance, whereas the safety risk prevention measures and safety accident handling measures correspond to the reasoning rules generated in Protégé 5.5.0. Originality/value This study proposes a new paradigm to improve the performance of safety management by expanding the derived value of SRs to establish relationships with safety risks and to respond effectively to combined risk characteristics.