A digital twin model for evidence-based clinical decision support in multiple myeloma treatment.

The treatment landscape for multiple myeloma (MM) has experienced substantial progress over the last decade. Despite the efficacy of new substances, patient responses tend to still be highly unpredictable. With increasing cognitive burden that is introduced through a complex and evolving treatment landscape, data-driven assistance tools are becoming more and more popular. Model-based approaches, such as digital twins (DT), enable simulation of probable responses to a set of input parameters based on retrospective observations. In the context of treatment decision-support, those mechanisms serve the goal to predict therapeutic outcomes to distinguish a favorable option from a potential failure. In the present work, we propose a similarity-based multiple myeloma digital twin (MMDT) that emphasizes explainability and interpretability in treatment outcome evaluation. We've conducted a requirement specification process using scientific literature from the medical and methodological domains to derive an architectural blueprint for the design and implementation of the MMDT. In a subsequent stage, we've implemented a four-layer concept where for each layer, we describe the utilized implementation procedure and interfaces to the surrounding DT environment. We further specify our solutions regarding the adoption of multi-line treatment strategies, the integration of external evidence and knowledge, as well as mechanisms to enable transparency in the data processing logic. Furthermore, we define an initial evaluation scenario in the context of patient characterization and treatment outcome simulation as an exemplary use case for our MMDT. Our derived MMDT instance is defined by 475 unique entities connected through 438 edges to form a MM knowledge graph. Using the MMRF CoMMpass real-world evidence database and a sample MM case, we processed a complete outcome assessment. The output shows a valid selection of potential treatment strategies for the integrated medical case and highlights the potential of the MMDT to be used for such applications. DT models face significant challenges in development, including availability of clinical data to algorithmically derive clinical decision support, as well as trustworthiness of the evaluated treatment options. We propose a collaborative approach that mitigates the regulatory and ethical concerns that are broadly discussed when automated decision-making tools are to be included into clinical routine.

Decision Support for Oropharyngeal Cancer Patients Based on Data-Driven Similarity Metrics for Medical Case Comparison.

Making complex medical decisions is becoming an increasingly challenging task due to the growing amount of available evidence to consider and the higher demand for personalized treatment and patient care. IT systems for the provision of clinical decision support (CDS) can provide sustainable relief if decisions are automatically evaluated and processed. In this paper, we propose an approach for quantifying similarity between new and previously recorded medical cases to enable significant knowledge transfer for reasoning tasks on a patient-level. Methodologically, 102 medical cases with oropharyngeal carcinoma were analyzed retrospectively. Based on independent disease characteristics, patient-specific data vectors including relevant information entities for primary and adjuvant treatment decisions were created. Utilizing the ϕK correlation coefficient as the methodological foundation of our approach, we were able to determine the predictive impact of each characteristic, thus enabling significant reduction of the feature space to allow for further analysis of the intra-variable distances between the respective feature states. The results revealed a significant feature-space reduction from initially 19 down to only 6 diagnostic variables (ϕK correlation coefficient ≥ 0.3, ϕK significance test ≥ 2.5) for the primary and 7 variables (from initially 14) for the adjuvant treatment setting. Further investigation on the resulting characteristics showed a non-linear behavior in relation to the corresponding distances on intra-variable level. Through the implementation of a 10-fold cross-validation procedure, we were further able to identify 8 (primary treatment) matching cases with an evaluation score of 1.0 and 9 (adjuvant treatment) matching cases with an evaluation score of 0.957 based on their shared treatment procedure as the endpoint for similarity definition. Based on those promising results, we conclude that our proposed method for using data-driven similarity measures for application in medical decision-making is able to offer valuable assistance for physicians. Furthermore, we consider our approach as universal in regard to other clinical use-cases, which would allow for an easy-to-implement adaptation for a range of further medical decision-making scenarios.

Clinical decision support models for oropharyngeal cancer treatment: design and evaluation of a multi-stage knowledge abstraction and formalization process.

PURPOSE: Treatment decisions in oncology are demanding and affect survival, general health, and quality of life. Expert systems can handle the complexity of the oncological field. We propose the application of a hybrid modeling approach for decision support models consisting of expert-based implementation of a decision model structure and machine-learning (ML) based parameter generation. We demonstrate our approach for the treatment of oropharyngeal cancer. METHODS: We created a clinical decision model based on Bayesian Networks and iteratively optimized its characteristics using structured knowledge engineering approaches. We combined manual adaptation of individual concepts with automatic learning of parameters and causalities. Using data from 94 patient records, we targeted the needed objectivity and clinical significance. RESULTS: In three iteration steps, we assessed the model with cross-validations. The initial aggregated accuracy of 0.529 could be increased to 0.883 in the final version. The predictive rates of the target nodes range from 0.557 to 0.960. CONCLUSION: Combining different methodological approaches requires balancing the complexity of the clinical subject matter with the amount of information available in the dataset for ML application. Our method showed promising results because flaws of one approach can be overcome by the other approach. However, technical integrability as well as clinical acceptance must always be ensured.

Bayesian Networks to Support Decision-Making for Immune-Checkpoint Blockade in Recurrent/Metastatic (R/M) Head and Neck Squamous Cell Carcinoma (HNSCC).

New diagnostic methods and novel therapeutic agents spawn additional and heterogeneous information, leading to an increasingly complex decision-making process for optimal treatment of cancer. A great amount of information is collected in organ-specific multidisciplinary tumor boards (MDTBs). By considering the patient's tumor properties, molecular pathological test results, and comorbidities, the MDTB has to consent an evidence-based treatment decision. Immunotherapies are increasingly important in today's cancer treatment, resulting in detailed information that influences the decision-making process. Clinical decision support systems can facilitate a better understanding via processing of multiple datasets of oncological cases and molecular genetic information, potentially fostering transparency and comprehensibility of available information, eventually leading to an optimum treatment decision for the individual patient. We constructed a digital patient model based on Bayesian networks to combine the relevant patient-specific and molecular data with depended probabilities derived from pertinent studies and clinical guidelines to calculate treatment decisions in head and neck squamous cell carcinoma (HNSCC). In a validation analysis, the model can provide guidance within the growing subject of immunotherapy in HNSCC and, based on its ability to calculate reliable probabilities, facilitates estimation of suitable therapy options. We compared actual treatment decisions of 25 patients with the calculated recommendations of our model and found significant concordance (Cohen's κ = 0.505, p = 0.009) and 84% accuracy.

Modeling and processing up-to-dateness of patient information in probabilistic therapy decision support.

OBJECTIVES: Probabilistic modeling of a patient's situation with the goal of providing calculated therapy recommendations can improve the decision making of interdisciplinary teams. Relevant information entities and direct causal dependencies, as well as uncertainty, must be formally described. Possible therapy options, tailored to the patient, can be inferred from the clinical data using these descriptions. However, there are several avoidable factors of uncertainty influencing the accuracy of the inference. For instance, inaccuracy may emerge from outdated information. In general, probabilistic models, e.g. Bayesian Networks can depict the causality and relations of individual information entities, but in general cannot evaluate individual entities concerning their up-to-dateness. The goal of the work at hand is to model diagnostic up-to-dateness, which can reasonably adjust the influence of outdated diagnostic information to improve the inference results of clinical decision models. METHODS AND MATERIALS: We analyzed 68 laryngeal cancer cases and modeled the state of up-to-dateness of different diagnostic modalities. All cases were used for cross-validation. 55 cases were used to train the model, 13 for testing. Each diagnostic procedure involved in the decision making process of these cases was associated with a specific threshold for the time the information is considered up-to-date, i.e. reliable. Based on this threshold, outdated findings could be identified and their impact on probabilistic calculations could be reduced. We applied the model for reducing the weight of outdated patient data in the computation of TNM stagings for the 13 test cases and compared the results to the manually derived TNM stagings in the patient files. RESULTS: With the implementation of these weights in the laryngeal cancer model, we increased the accuracy of the TNM calculation from 0.61 (8 out of 13 cases correct) to 0.76 (10 out of 13 cases correct). CONCLUSION: Decision delay may cause specific patient data to be outdated. This can cause contradictory or false information and impair calculations for clinical decision support. Our approach demonstrates that the accuracy of Bayesian Network models can be improved when pre-processing the patient-specific data and evaluating their up-to-dateness with reduced weights on outdated information.

Modular Architecture for Integrated Model-Based Decision Support.

Model-based decision support systems promise to be a valuable addition to oncological treatments and the implementation of personalized therapies. For the integration and sharing of decision models, the involved systems must be able to communicate with each other. In this paper, we propose a modularized architecture of dedicated systems for the integration of probabilistic decision models into existing hospital environments. These systems interconnect via web services and provide model sharing and processing capabilities for clinical information systems. Along the lines of IHE integration profiles from other disciplines and the meaningful reuse of routinely recorded patient data, our approach aims for the seamless integration of decision models into hospital infrastructure and the physicians' daily work.

Information architecture for a patient-specific dashboard in head and neck tumor boards.

PURPOSE: Overcoming the flaws of current data management conditions in head and neck oncology could enable integrated information systems specifically tailored to the needs of medical experts in a tumor board meeting. Clinical dashboards are a promising method to assist various aspects of the decision-making process in such cognitively demanding scenarios. However, in order to provide extensive and intuitive assistance to the participating physicians, the design and development of such a system have to be user-centric. To accomplish this task, conceptual methods need to be performed prior to the technical development and integration stages. METHODS: We have conducted a qualitative survey including eight clinical experts with different levels of expertise in the field of head and neck oncology. According to the principles of information architecture, the survey focused on the identification and causal interconnection of necessary metrics for information assessment in the tumor board. RESULTS: Based on the feedback by the clinical experts, we have constructed a detailed map of the required information items for a tumor board dashboard in head and neck oncology. Furthermore, we have identified three distinct groups of metrics (patient, disease and therapy metrics) as well as specific recommendations for their structural and graphical implementation. CONCLUSION: By using the information architecture, we were able to gather valuable feedback about the requirements and cognitive processes of the tumor board members. Those insights have helped us to develop a dashboard application that closely adapts to the specified needs and characteristics, and thus is primarily user-centric.