Personalized Decision Support Research Articles

Personalised decision support in the management of patients with musculoskeletal pain in primary physiotherapy care: a cluster randomised controlled trial (the SupportPrim project).

We developed the SupportPrim PT clinical decision support system (CDSS) using the artificial intelligence method case-based reasoning to support personalised musculoskeletal pain management. The aim of this study was to evaluate the effectiveness of the CDSS for patients in physiotherapy practice. A cluster randomised controlled trial was conducted in primary care in Norway. We randomised 44 physiotherapists to (1) use the CDSS alongside usual care or (2) usual care alone. The CDSS provided personalised treatment recommendations based on a case base of 105 patients with positive outcomes. During the trial, the case-based reasoning system did not have an active learning capability; therefore, the case base size remained the same throughout the study. We included 724 patients presenting with neck, shoulder, back, hip, knee, or complex pain (CDSS; n = 358, usual care; n = 366). Primary outcomes were assessed with multilevel logistic regression using self-reported Global Perceived Effect (GPE) and Patient-Specific Functional Scale (PSFS). At 12 weeks, 165/298 (55.4%) patients in the intervention group and 176/321 (54.8%) in the control group reported improvement in GPE (odds ratio, 1.18; confidence interval, 0.50-2.78). For PSFS, 173/290 (59.7%) patients in the intervention group and 218/310 (70.3%) in the control group reported clinically important improvement in function (odds ratio, 0.41; confidence interval, 0.20-0.85). No significant between-group differences were found for GPE. For PSFS, there was a significant difference favouring the control group, but this was less than the prespecified difference of 15%. We identified several study limitations and recommend further investigation into artificial intelligence applications for managing musculoskeletal pain.

A generalizable normative deep autoencoder for brain morphological anomaly detection: application to the multi-site StratiBip dataset on bipolar disorder in an external validation framework.

The heterogeneity of psychiatric disorders makes researching disorder-specific neurobiological markers an ill-posed problem. Here, we face the need for disease stratification models by presenting a generalizable multivariate normative modelling framework for characterizing brain morphology, applied to bipolar disorder (BD). We employed deep autoencoders in an anomaly detection framework, combined with a confounder removal step integrating training and external validation. The model was trained with healthy control (HC) data from the human connectome project and applied to multi-site external data of HC and BD individuals. We found that brain deviating scores were greater, more heterogeneous, and with increased extreme values in the BD group, with volumes prominently from the basal ganglia, hippocampus and adjacent regions emerging as significantly deviating. Similarly, individual brain deviating maps based on modified z scores expressed higher abnormalities occurrences, but their overall spatial overlap was lower compared to HCs. Our generalizable framework enabled the identification of subject- and group-level brain normative-deviating patterns, a step forward towards the development of more effective and personalized clinical decision support systems and patient stratification in psychiatry.

Prediction of Microinvasion in Breast Ductal Carcinoma in Situ Using Conventional Ultrasound Combined with Contrast-Enhanced Ultrasound Features: A Two-Center Study

BackgroundTo develop and validate a model based on conventional ultrasound (CUS) and contrast-enhanced ultrasound (CEUS) features to preoperatively predict microinvasion in breast ductal carcinoma in situ (DCIS). Patients and MethodsData from 163 patients with DCIS who underwent CUS and CEUS from the internal hospital was retrospectively collected and randomly apportioned into training and internal validation sets in a ratio of 7:3. External validation set included 56 patients with DCIS from the external hospital. Univariate and multivariate logistic regression analysis were performed to determine the independent risk factors associated with microinvasion. These factors were used to develop predictive models. The performance was evaluated through calibration, discrimination, and clinical utility. ResultsMultivariate analysis indicated that centripetal enhancement direction (odds ratio [OR], 13.268; 95% confidence interval [CI], 3.687-47.746) and enhancement range enlarged on CEUS (OR, 4.876; 95% CI, 1.470-16.181), lesion size of ≥20 mm (OR, 3.265; 95% CI, 1.230-8.669) and calcification detected on CUS (OR, 5.174; 95% CI, 1.903-14.066) were independent risk factors associated with microinvasion. The nomogram incorporated the CUS and CEUS features achieved favorable discrimination (AUCs of 0.850, 0.848, and 0.879 for the training, internal and external validation datasets), with good calibration. The nomogram outperformed the CUS model and CEUS model (all P < .05). Decision curve analysis confirmed that the predictive nomogram was clinically useful. ConclusionThe nomogram based on CUS and CEUS features showed promising predictive value for the preoperative identification of microinvasion in patients with DCIS.

Improving Healthcare Quality with a LHS: From Patient-Generated Health Data to Evidence-Based Recommendations.

One approach to enriching the Learning Health System (LHS) is leveraging vital signs and data from wearable technologies. Blood oxygen, heart rate, respiration rates, and other data collected by wearables (like sleep and exercise patterns) can be used to monitor and predict health conditions. This data is already being collected and could be used to improve healthcare in several ways. Our approach will be health data interoperability with HL7 FHIR (for data exchange between different systems), openEHR (to store researchable data separated from software but connected to ontologies, external terminologies and code sets) and maintain the semantics of data. OpenEHR is a standard that has an important role in modelling processes and clinical decisions. The six pillars of Lifestyle Medicine can be a first attempt to change how patients see their daily decisions, affecting the mid to long-term evolution of their health. Our objective is to develop the first stage of the LHS based on a co-produced personal health recording (CoPHR) built on top of a local LLM that interoperates health data through HL7 FHIR, openEHR, OHDSI and terminologies that can ingest external evidence and produces clinical and personal decision support and, when combined with many other patients, can produce or confirm evidence.

Joint AI-driven event prediction and longitudinal modeling in newly diagnosed and relapsed multiple myeloma

Multiple myeloma management requires a balance between maximizing survival, minimizing adverse events to therapy, and monitoring disease progression. While previous work has proposed data-driven models for individual tasks, these approaches fail to provide a holistic view of a patient’s disease state, limiting their utility to assist physician decision-making. To address this limitation, we developed a transformer-based machine learning model that jointly (1) predicts progression-free survival (PFS), overall survival (OS), and adverse events (AE), (2) forecasts key disease biomarkers, and (3) assesses the effect of different treatment strategies, e.g., ixazomib, lenalidomide, dexamethasone (IRd) vs lenalidomide, dexamethasone (Rd). Using TOURMALINE trial data, we trained and internally validated our model on newly diagnosed myeloma patients (N = 703) and externally validated it on relapsed and refractory myeloma patients (N = 720). Our model achieved superior performance to a risk model based on the multiple myeloma international staging system (ISS) (p < 0.001, Bonferroni corrected) and comparable performance to survival models trained separately on each task, but unable to forecast biomarkers. Our approach outperformed state-of-the-art deep learning models, tailored towards forecasting, on predicting key disease biomarkers (p < 0.001, Bonferroni corrected). Finally, leveraging our model’s capacity to estimate individual-level treatment effects, we found that patients with IgA kappa myeloma appear to benefit the most from IRd. Our study suggests that a holistic assessment of a patient’s myeloma course is possible, potentially serving as the foundation for a personalized decision support system.

Should AI-Powered Whole-Genome Sequencing Be Used Routinely for Personalized Decision Support in Surgical Oncology—A Scoping Review

In this scoping review, we delve into the transformative potential of artificial intelligence (AI) in addressing challenges inherent in whole-genome sequencing (WGS) analysis, with a specific focus on its implications in oncology. Unveiling the limitations of existing sequencing technologies, the review illuminates how AI-powered methods emerge as innovative solutions to surmount these obstacles. The evolution of DNA sequencing technologies, progressing from Sanger sequencing to next-generation sequencing, sets the backdrop for AI’s emergence as a potent ally in processing and analyzing the voluminous genomic data generated. Particularly, deep learning methods play a pivotal role in extracting knowledge and discerning patterns from the vast landscape of genomic information. In the context of oncology, AI-powered methods exhibit considerable potential across diverse facets of WGS analysis, including variant calling, structural variation identification, and pharmacogenomic analysis. This review underscores the significance of multimodal approaches in diagnoses and therapies, highlighting the importance of ongoing research and development in AI-powered WGS techniques. Integrating AI into the analytical framework empowers scientists and clinicians to unravel the intricate interplay of genomics within the realm of multi-omics research, paving the way for more successful personalized and targeted treatments.

Personalized decision support system for tailoring IgA nephropathy treatment strategies

BackgroundThe ongoing debate surrounding the use of immunosuppressive treatments for IgA nephropathy (IgAN) underscores the demand for personalized and effective strategies. MethodsAnalyzed data from 807 IgAN patients over 5+ years using three methods: Random Forest with molecular biomarkers, network biomarkers with graph engineering, and an auto-encoder model. All models were trained using identical demographic, clinical, and pathological data, employing an 80–20 split for training and testing purposes. ResultsIn the comprehensive assessment of IgAN prognosis, the Random Forest model, employing molecular biomarkers, demonstrated strong performance metrics (AUC = 0.83, sensitivity = 0.51, specificity = 0.96). However, traditional graph feature engineering on patient-specific networks outperformed these results with an AUC of 0.90, sensitivity of 0.64, and specificity of 0.94. The Auto-encoder model showed the best accuracy (AUC = 0.91, sensitivity = 0.46, specificity = 0.96). The findings highlighted the superior predictive capabilities of network biomarkers over molecular biomarkers for adverse renal outcome prediction in IgAN. Consequently, we integrated Auto-encoder-derived Network Biomarkers with Random Forest Models to enhance prognostic precision in diverse IgAN treatment scenarios. The prediction for the prognosis of patients receiving supportive care, glucocorticoid therapy, and immunosuppressant treatment yielded AUC values of 0.95, 0.96, and 1, respectively, indicating high specificity. Drawing from these insights, we pioneered the development of an innovative decision support model for IgAN treatment. This model demonstrated the ability to make medical decisions comparable to those by experienced nephrologists, enabling the customization of personalized disease management strategies. ConclusionOur system accurately predicted IgAN prognosis and evaluated various treatment efficacies, aiding physicians in devising optimal therapeutic strategies for patients.

A text analytics approach for mining public discussions in online cancer forum: Analysis of multi-intent lung cancer treatment dataset

BackgroundOnline cancer forums (OCF) are increasingly popular platforms for patients and caregivers to discuss, seek information on, and share opinions about diseases and treatments. This interaction generates a substantial amount of unstructured text data, necessitating deeper exploration. Using time series data, our study exploits topic modeling in the novel domain of online cancer forums (OCFs) to identify meaningful topics and changing dynamics of online discussion across different lung cancer treatment intent groups. MethodsFor this purpose, a dataset comprising 27,998 forum posts about lung cancer was collected from three OCFs: lungcancer.net, lungevity.org, and reddit.com, spanning the years 2016 to 2018. ResultsThe analysis reflects the public discussion on multi-intent lung cancer treatment over time, taking into account seasonal variations. Discussions on cancer symptoms and prevention garnered the most attention, dominating both curative and palliative care discussions. There were distinct seasonal peaks: curative care topics surged from winter to late spring, while palliative care topics peaked from late summer to mid-autumn. Keyword analysis highlighted that lung cancer diagnosis and treatment were primary topics, whereas cancer prevention and treatment outcomes were predominant across multi-care contexts. For the study period, curative care discussions predominantly revolved around informational support and disease syndromes. In contrast, social support and cancer prevention prevailed in the palliative care context. Notably, topics such as cancer screening and cancer treatment exhibit pronounced seasonal variations in curative care, peaking in frequency during the summers (May to August) of the study period. Meanwhile, the topic of tumor control within palliative care showed significant seasonal influence during the winters and summers of 2017 and 2018. ConclusionOur text analysis approach using OCF data shows potential for computational methods in this novel domain to gain insights into trends in public cancer communication and seasonal variations for a better understanding of improving personalized care, decision support, treatment outcomes, and quality of life.

Dynamic Prognosis Prediction for Patients on DAPT After Drug-Eluting Stent Implantation: Model Development and Validation.

The rapid evolution of artificial intelligence (AI) in conjunction with recent updates in dual antiplatelet therapy (DAPT) management guidelines emphasizes the necessity for innovative models to predict ischemic or bleeding events after drug-eluting stent implantation. Leveraging AI for dynamic prediction has the potential to revolutionize risk stratification and provide personalized decision support for DAPT management. We developed and validated a new AI-based pipeline using retrospective data of drug-eluting stent-treated patients, sourced from the Cerner Health Facts data set (n=98 236) and Optum's de-identified Clinformatics Data Mart Database (n=9978). The 36 months following drug-eluting stent implantation were designated as our primary forecasting interval, further segmented into 6 sequential prediction windows. We evaluated 5 distinct AI algorithms for their precision in predicting ischemic and bleeding risks. Model discriminative accuracy was assessed using the area under the receiver operating characteristic curve, among other metrics. The weighted light gradient boosting machine stood out as the preeminent model, thus earning its place as our AI-DAPT model. The AI-DAPT demonstrated peak accuracy in the 30 to 36 months window, charting an area under the receiver operating characteristic curve of 90% [95% CI, 88%-92%] for ischemia and 84% [95% CI, 82%-87%] for bleeding predictions. Our AI-DAPT excels in formulating iterative, refined dynamic predictions by assimilating ongoing updates from patients' clinical profiles, holding value as a novel smart clinical tool to facilitate optimal DAPT duration management with high accuracy and adaptability.

Utilizing RoBERTa Model for Churn Prediction through Clustered Contextual Conversation Opinion Mining

In computational study and automatic recognition of opinions in free texts, certain words in sentences are used to decide its sentiments. While analysing each customer’s opinion per time in churn management will be effective for personalised recommendations. Oftentimes, the opinion is not sufficient for contextualised content mining. While personalised recommendations are time consuming, it also does not provide complete picture of an overall sentiment in the business community of customers. To help businesses identify widespread issues affecting a large segment of their customers towards engendering patterns and trends of different customer churn behaviour, here, we developed a clustered contextualised conversation as opinions set for integration with Roberta Model. The developed churn behavioural opinion clusters disambiguated short messages while charactering contents collectively based on context beyond keyword-based sentiment matching for effective mining. Based on the predicted opinion threshold, customer churn category for group-based personalised decision support was generated, with matching concepts. The baseline RoBERTa model on the contextually clustered opinions, trained with a batch size of 16, a learning rate of 2e-5, over 8 epochs, using a maximum sequence length of 128 and standard hyperparameters, achieved an accuracy of 92%, Precision of 88%, Recall of 86% and F1 score of 84% over a test set of 30%.

Semantic Integration of BPMN Models and FHIR Data to Enable Personalized Decision Support for Malignant Melanoma

With digital patient data increasing due to new diagnostic methods and technology, showing the right data in the context of decision support at the point of care becomes an even greater challenge. Standard operating procedures (SOPs) modeled in BPMN (Business Process Model and Notation) contain evidence-based treatment guidance for all phases of a certain diagnosis, while physicians need the parts relevant to a specific patient at a specific point in the clinical process. Therefore, integration of patient data from electronic health records (EHRs) providing context to clinicians is needed, which is stored and communicated in HL7 (Health Level Seven) FHIR (Fast Healthcare Interoperability Resources). To address this issue, we propose a method combining an integration of stored data into BPMN and a loss-free transformation from BPMN into FHIR, and vice versa. Based on that method, an identification of the next necessary decision point in a specific patient context is possible. We verified the method for treatment of malignant melanoma by using an extract of a formalized SOP document with predefined decision points and validated FHIR references with real EHR data. The patient data could be stored and integrated into the BPMN element ‘DataStoreReference’. Our loss-free transformation process therefore is the foundation for combining evidence-based knowledge from formalized clinical guidelines or SOPs and patient data from EHRs stored in FHIR. Processing the SOP with the available patient data can then lead to the next upcoming decision point, which will be displayed to the physician integrated with the corresponding data.

Revolutionizing Patient Care: A Comprehensive Review of Artificial Intelligence Applications in Anesthesia.

This review explores the intersection of artificial intelligence (AI) and anesthesia, examining its transformative impact on patient care across various phases. Beginning with a historical overview of anesthesia, we highlight the critical role of technological advancements in ensuring optimal patient outcomes. The emergence of AI in healthcare sets the stage for a comprehensive analysis of its applications in anesthesia. In the preoperative phase, AI facilitates personalized risk assessments and decision support, optimizing anesthesia planning and drug dosage predictions. Moving to the intraoperative phase, we delve into AI's role in monitoring and control through sophisticated anesthesia monitoring and closed-loop systems. Additionally, we discuss the integration of robotics and AI-guided procedures, revolutionizing surgical assistance. Transitioning to the postoperative phase, we explore AI-driven postoperative monitoring, predictive analysis for complications, and the integration of AI into rehabilitation programs and long-term follow-up. These new applications redefine patient recovery, emphasizing personalized care and proactive interventions. However, the integration of AI in anesthesia poses challenges and ethical considerations. Data security, interpretability, and bias in AI algorithms demand scrutiny. Moreover, the evolving patient-doctor relationship in an AI-driven care landscape requires a delicate balance between efficiency and human touch. Looking forward, we discuss the future directions of AI in anesthesia, anticipating advances in technology and AI algorithms. The integration of AI into routine clinical practice and its potential impact on anesthesia education and training are explored, emphasizing the need for collaboration, education, and ethical guidelines. This review provides a comprehensive overview of AI applications in anesthesia, offering insights into the present landscape, challenges, and future directions. The synthesis of historical perspectives, current applications, and future possibilities underscores the transformative potential of AI in revolutionizing patient care within the dynamic field of anesthesia.

From E-Commerce to Virtual Commerce: The Exciting Opportunities of Virtual Shopping

Abstract Virtual commerce has the potential to transform the shopping experience. It is highly interactive and can create immersive experiences when consumers wear head-mounted displays. Benefits include increased telepresence, better size and proportion judgement, customizable shopping environments, highly personalized decision support and social interaction opportunities. Technological advances in VR hardware, such as improved display resolution and reduced motion sickness, contribute to a more seamless and enjoyable shopping experience. These benefits can lead to higher purchase rates. However, there are also privacy concerns due to the many ways in which data can be collected. While the future of VR shopping is difficult to predict, AR applications are already well established. Tech giants, research institutions, and startups will play a critical role in shaping the future of virtual commerce.

Personalized Federated Learning for Institutional Prediction Model using Electronic Health Records: A Covariate Adjustment Approach.

Federated learning (FL) has attracted attention as a technology that allows multiple medical institutions to collaborate on AI without disclosing each other's patient data. However, FL has the challenge of being unable to robustly learn when the data of participating clients is non-independently and non-identically distributed (Non-IID). Personalized Federated Learning (PFL), which constructs a personalized model for each client, has been proposed as a solution to this problem. However, conventional PFL methods do not ensure the interpretability of personalization, specifically, the identification of which data samples are contributed to each personalized learning, which is important for AI in medical applications. In this study, we propose a novel PFL framework, Federated Adjustment of Covariate (FedCov), which acquires a propensity score model representing the covariate shift among clients through prior FL, then learns a final model by weighting the contribution of each training sample to PFL based on the estimated propensity score. This approach enables both the learning of personalized models through covariate adjustment and the visualization of the contribution of each client to PFL. FedCov was evaluated in the prediction of in-hospital mortality across 50 hospitals in the eICU Collaborative Research Database, achieving an ROC-AUC of 0.750. This result outperformed the AUCs in the 0.720-0.735 range achieved by conventional FL methods and was closest to the AUC of 0.754 achieved by centralized learning.Clinical Relevance- This study demonstrates the feasibility of providing sophisticated and personalized AI-driven clinical decision support to any medical institution through personalized federated learning.

Deep learning model to predict exercise stress test results: Optimizing the diagnostic test selection strategy and reduce wastage in suspected coronary artery disease patients

BackgroundCardiac exercise stress testing (EST) offers a non-invasive way in the management of patients with suspected coronary artery disease (CAD). However, up to 30% EST results are either inconclusive or non-diagnostic, which results in significant resource wastage. Our aim was to build machine learning (ML) based models, using patients demographic (age, sex) and pre-test clinical information (reason for performing test, medications, blood pressure, heart rate, and resting electrocardiogram), capable of predicting EST results beforehand including those with inconclusive or non-diagnostic results. MethodsA total of 30,710 patients (mean age 54.0 years, 69% male) were included in the study with 25% randomly sampled in the test set, and the remaining samples were split into a train and validation set with a ratio of 9:1. We constructed different ML models from pre-test variables and compared their discriminant power using the area under the receiver operating characteristic curve (AUC). ResultsA network of Oblivious Decision Trees provided the best discriminant power (AUC=0.83, sensitivity=69%, specificity=0.78%) for predicting inconclusive EST results. A total of 2010 inconclusive ESTs were correctly identified in the testing set. ConclusionsOur ML model, developed using demographic and pre-test clinical information, can accurately predict EST results and could be used to identify patients with inconclusive or non-diagnostic results beforehand. Our system could thus be used as a personalised decision support tool by clinicians for optimizing the diagnostic test selection strategy for CAD patients and to reduce healthcare expenditure by reducing nondiagnostic or inconclusive ESTs.

Predicting favorable response to intravenous morphine in pediatric critically ill cardiac patients.

Analgesia and sedation are integral to the care of critically ill children. However, the choice and dose of the analgesic or sedative drug is often empiric, and models predicting favorable responses are lacking. We aimed to compute models to predict a patient's response to intravenous morphine. We retrospectively analyzed data from consecutive patients admitted to the Cardiac Intensive Care Unit (January 2011-January 2020) who received at least one intravenous bolus of morphine. The primary outcome was a decrease in the State Behavioral Scale (SBS) ≥1 point; the secondary outcome was a decrease in the heart rate Z-score (zHR) at 30 min. Effective doses were modeled using logistic regression, Lasso regression, and random forest modeling. A total of 117,495 administrations of intravenous morphine among 8140 patients (median age 0.6 years [interquartile range [IQR] 0.19, 3.3]) were included. The median morphine dose was 0.051 mg/kg (IQR 0.048, 0.099) and the median 30-day cumulative dose was 2.2 mg/kg (IQR 0.4, 15.3). SBS decreased following 30% of doses, did not change following 45%, and increased following 25%. The zHR significantly decreased after morphine administration (median delta-zHR -0.34 [IQR-1.03, 0.00], p < 0.001). The following factors were associated with favorable response to morphine: A concomitant infusion of propofol, higher prior 30-day cumulative dose, being invasively ventilated and/or on vasopressors. Higher morphine dose, higher zHR pre-morphine, an additional analgosedation bolus ±30 min around the index bolus, a concomitant ketamine or dexmedetomidine infusion, and showing signs of withdrawal syndrome were associated with unfavorable response. Logistic regression (area under the receiver operating characteristic [ROC] curve [AUC] 0.900) and machine learning models (AUC 0.906) performed comparably, with a sensitivity of 95%, specificity of 71%, and negative predictive value of 97%. Statistical models identify 95% of effective intravenous morphine doses in pediatric critically ill cardiac patients, while incorrectly suggesting an effective dose in 29% of cases. This work represents an important step toward computer-aided, personalized clinical decision support tool for sedation and analgesia in ICU patients.

Digitizing local search: An empirical analysis of mobile search behavior in offline shopping

Mobile devices have become increasingly prevalent in the context of consumers' offline search and shopping behavior by acting as personal decision support tools. This paper investigates the relationship between consumers' real-time location and their search behavior when shopping offline. We develop two location-specific measures based on mobile device usage to analyze the association between a consumer's real-time location and their search behavior: Present store distance and present store density. We apply these measures to a unique dataset from a leading mobile product information app that includes >13 million search sessions from 2.5 million consumers searching for 1.8 million different products. Our findings show that store distance and store density significantly influence consumers' mobile search behavior: detailed searches for product-specific attributes (search depth) increase as store distance decreases and store density increases, while the number of jointly searched products within the same category (search breadth) increases as store distance and density increase. We also find that durable products weaken the main effects of store distance and density on search depth but strengthen it for search breadth. These results provide insights for offline retailers to better understand location-specific consumer demands and improve their mobile in-app targeting and product assortment decisions.

QuantImage v2: a comprehensive and integrated physician-centered cloud platform for radiomics and machine learning research

BackgroundRadiomics, the field of image-based computational medical biomarker research, has experienced rapid growth over the past decade due to its potential to revolutionize the development of personalized decision support models. However, despite its research momentum and important advances toward methodological standardization, the translation of radiomics prediction models into clinical practice only progresses slowly. The lack of physicians leading the development of radiomics models and insufficient integration of radiomics tools in the clinical workflow contributes to this slow uptake.MethodsWe propose a physician-centered vision of radiomics research and derive minimal functional requirements for radiomics research software to support this vision. Free-to-access radiomics tools and frameworks were reviewed to identify best practices and reveal the shortcomings of existing software solutions to optimally support physician-driven radiomics research in a clinical environment.ResultsSupport for user-friendly development and evaluation of radiomics prediction models via machine learning was found to be missing in most tools. QuantImage v2 (QI2) was designed and implemented to address these shortcomings. QI2 relies on well-established existing tools and open-source libraries to realize and concretely demonstrate the potential of a one-stop tool for physician-driven radiomics research. It provides web-based access to cohort management, feature extraction, and visualization and supports “no-code” development and evaluation of machine learning models against patient-specific outcome data.ConclusionsQI2 fills a gap in the radiomics software landscape by enabling “no-code” radiomics research, including model validation, in a clinical environment. Further information about QI2, a public instance of the system, and its source code is available at https://medgift.github.io/quantimage-v2-info/.Key pointsAs domain experts, physicians play a key role in the development of radiomics models.Existing software solutions do not support physician-driven research optimally.QuantImage v2 implements a physician-centered vision for radiomics research.QuantImage v2 is a web-based, “no-code” radiomics research platform.

Partially observable collaborative model for optimizing personalized treatment selection

Precision medicine that enables personalized treatment decision support has become an increasingly important research topic in chronic disease care. The main challenges in designing a treatment algorithm include modeling individual disease progression dynamics and designing adaptive treatment selection strategy. This study aims to develop an adaptive treatment selection framework tailored to an individual patient’s disease progression pattern and treatment response. We propose a Partially Observable Collaborative Model (POCM) to capture the individual variations in a heterogeneous population and optimize treatment outcomes in three stages. The POCM first infers the disease progression models by subgroup patterns using population data in stage one and then fine-tunes the models for individual patients with a small number of treatment trials in stage two. In stage three, we show how the treatment policies based on the Partially Observable Markov Decision Process (POMDP) can be tailored to individual patients by utilizing the disease models learned from the POCM. Using a simulated population of chronic depression patients, we show that the POCM can more accurately estimate the personal disease progression than the traditional method of solving a hidden Markov model. We also compare the POMDP treatment policies with other heuristic policies and demonstrate that the POCM-based policies give the highest net monetary benefits in majority of parameter settings. To conclude, the POCM method is a promising approach to model the chronic disease progression process and recommend a personalized treatment plan for individual patients in a heterogeneous population.

Real-world performance analysis of a novel computational method in the precision oncology of pediatric tumors.

The utility of routine extensive molecular profiling of pediatric tumors is a matter of debate due to the high number of genetic alterations of unknown significance or low evidence and the lack of standardized and personalized decision support methods. Digital drug assignment (DDA) is a novel computational method to prioritize treatment options by aggregating numerous evidence-based associations between multiple drivers, targets, and targeted agents. DDA has been validated to improve personalized treatment decisions based on the outcome data of adult patients treated in the SHIVA01 clinical trial. The aim of this study was to evaluate the utility of DDA in pediatric oncology. Between 2017 and 2020, 103 high-risk pediatric cancer patients (< 21years) were involved in our precision oncology program, and samples from 100 patients were eligible for further analysis. Tissue or blood samples were analyzed by whole-exome (WES) or targeted panel sequencing and other molecular diagnostic modalities and processed by a software system using the DDA algorithm for therapeutic decision support. Finally, a molecular tumor board (MTB) evaluated the results to provide therapy recommendations. Of the 100 cases with comprehensive molecular diagnostic data, 88 yielded WES and 12 panel sequencing results. DDA identified matching off-label targeted treatment options (actionability) in 72/100 cases (72%), while 57/100 (57%) showed potential drug resistance. Actionability reached 88% (29/33) by 2020 due to the continuous updates of the evidence database. MTB approved the clinical use of a DDA-top-listed treatment in 56 of 72 actionable cases (78%). The approved therapies had significantly higher aggregated evidence levels (AELs) than dismissed therapies. Filtering of WES results for targeted panels missed important mutations affecting therapy selection. DDA is a promising approach to overcome challenges associated with the interpretation of extensive molecular profiling in the routine care of high-risk pediatric cancers. Knowledgebase updates enable automatic interpretation of a continuously expanding gene set, a "virtual" panel, filtered out from genome-wide analysis to always maximize the performance of precision treatment planning.