Search

Abstract Background Artificial intelligence (AI) has become a pivotal tool in advancing contemporary personalised medicine, with the goal of tailoring treatments to individual patient conditions. This has heightened the demand for access to diverse data from clinical practice and daily life for research, posing challenges due to the sensitive nature of medical information, including genetics and health conditions. Regulations like the Health Insurance Portability and Accountability Act (HIPAA) in the U.S. and the General Data Protection Regulation (GDPR) in Europe aim to strike a balance between data security, privacy, and the imperative for access. Results We present the Gemelli Generator - Real World Data (GEN-RWD) Sandbox, a modular multi-agent platform designed for distributed analytics in healthcare. Its primary objective is to empower external researchers to leverage hospital data while upholding privacy and ownership, obviating the need for direct data sharing. Docker compatibility adds an extra layer of flexibility, and scalability is assured through modular design, facilitating combinations of Proxy and Processor modules with various graphical interfaces. Security and reliability are reinforced through components like Identity and Access Management (IAM) agent, and a Blockchain-based notarisation module. Certification processes verify the identities of information senders and receivers. Conclusions The GEN-RWD Sandbox architecture achieves a good level of usability while ensuring a blend of flexibility, scalability, and security. Featuring a user-friendly graphical interface catering to diverse technical expertise, its external accessibility enables personnel outside the hospital to use the platform. Overall, the GEN-RWD Sandbox emerges as a comprehensive solution for healthcare distributed analytics, maintaining a delicate equilibrium between accessibility, scalability, and security.

Abstract Predictive modelling of cancer outcomes using radiomics faces dimensionality problems and data limitations, as radiomics features often number in the hundreds, and multi-institutional data sharing is ()often unfeasible. Federated learning (FL) and feature selection (FS) techniques combined can help overcome these issues, as one provides the means of training models without exchanging sensitive data, while the other identifies the most informative features, reduces overfitting, and improves model interpretability. Our proposed FS pipeline based on FL principles targets data-driven radiomics FS in a multivariate survival study of non-small cell lung cancer patients. The pipeline was run across datasets from three institutions without patient-level data exchange. It includes two FS techniques, Correlation-based Feature Selection and LASSO regularization, and Cox Proportional-Hazard regression with Overall Survival as endpoint. Trained and validated on 828 patients overall, our pipeline yielded a radiomic signature comprising "intensity-based energy" and "mean discretised intensity". Validation resulted in a mean Harrell C-index of 0.59, showcasing fair efficacy in risk stratification. In conclusion, we suggest a distributed radiomics approach that incorporates preliminary feature selection to systematically decrease the feature set based on data-driven considerations. This aims to address dimensionality challenges beyond those associated with data constraints and interpretability concerns.

The aim of the study was to build a machine learning-based predictive model to discriminate between hospitalized patients at low risk and high risk of bloodstream infection (BSI). A Data Mart including all patients hospitalized between January 2016 and December 2019 with suspected BSI was built. Multivariate logistic regression was applied to develop a clinically interpretable machine learning predictive model. The model was trained on 2016–2018 data and tested on 2019 data. A feature selection based on a univariate logistic regression first selected candidate predictors of BSI. A multivariate logistic regression with stepwise feature selection in five-fold cross-validation was applied to express the risk of BSI. A total of 5660 hospitalizations (4026 and 1634 in the training and the validation subsets, respectively) were included. Eleven predictors of BSI were identified. The performance of the model in terms of AUROC was 0.74. Based on the interquartile predicted risk score, 508 (31.1%) patients were defined as being at low risk, 776 (47.5%) at medium risk, and 350 (21.4%) at high risk of BSI. Of them, 14.2% (72/508), 30.8% (239/776), and 64% (224/350) had a BSI, respectively. The performance of the predictive model of BSI is promising. Computational infrastructure and machine learning models can help clinicians identify people at low risk for BSI, ultimately supporting an antibiotic stewardship approach.

Background and Purpose: Tumor recurrence, a characteristic of malignant tumors, is the biggest concern for rectal cancer survivors. The epidemiology of the disease calls for a pressing need to improve healthcare quality and patient outcomes. Prediction models such as Bayesian networks, which can probabilistically reason under uncertainty, could assist caregivers with patient management. However, some concerns are associated with the standard approaches to developing these structures in medicine. Therefore, this study aims to compare Bayesian network structures that stem from these two techniques. Materials and Methods: A retrospective analysis was performed on 6754 locally advanced rectal cancer (LARC) patients enrolled in 14 international clinical trials. Local tumor recurrence at 2, 3, and 5-years was defined as the endpoints of interest. Five rectal cancer treating physicians from three countries elicited the expert structure. The algorithmic structure was inferred from the data with the hill-climbing algorithm. Structural performance was assessed with calibration plots and area under the curve values. Results: The area under the curve for the expert structure on the training and validation data was above 0.9 and 0.8, respectively, for all the time points. However, the algorithmic structure had superior predictive performance over the expert structure for all time points of interest. Conclusion: We have developed and internally validated a Bayesian networks structure from experts’ opinions, which can predict the risk of a LARC patient developing a tumor recurrence at 2, 3, and 5 years. Our result shows that the algorithmic-based structures are more performant and less interpretable than expert-based structures.

In the era of evidence-based medicine, several clinical guidelines were developed, supporting cancer management from diagnosis to treatment and aiming to optimize patient care and hospital resources. Nevertheless, individual patient characteristics and organizational factors may lead to deviations from these standard recommendations during clinical practice. In this context, process mining in healthcare constitutes a valid tool to evaluate conformance of real treatment pathways, extracted from hospital data warehouses as event log, to standard clinical guidelines, translated into computer-interpretable formats. In this study we translate the European Society of Medical Oncology guidelines for rectal cancer treatment into a computer-interpretable format using Pseudo-Workflow formalism (PWF), a language already employed in pMineR software library for Process Mining in Healthcare. We investigate the adherence of a real-world cohort of rectal cancer patients treated at Fondazione Policlinico Universitario A. Gemelli IRCCS, data associated with cancer diagnosis and treatment are extracted from hospital databases in 453 patients diagnosed with rectal cancer. PWF enables the easy implementation of guidelines in a computer-interpretable format and visualizations that can improve understandability and interpretability of physicians. Results of the conformance checking analysis on our cohort identify a subgroup of patients receiving a long course treatment that deviates from guidelines due to a moderate increase in radiotherapy dose and an addition of oxaliplatin during chemotherapy treatment. This study demonstrates the importance of PWF to evaluate clinical guidelines adherence and to identify reasons of deviations during a treatment process in a real-world and multidisciplinary setting.

BackgroundHeart failure (HF) is a multifaceted clinical syndrome characterized by different etiologies, risk factors, comorbidities, and a heterogeneous clinical course. The current model, based on data from clinical trials, is limited by the biases related to a highly-selected sample in a protected environment, constraining the applicability of evidence in the real-world scenario. If properly leveraged, the enormous amount of data from real-world may have a groundbreaking impact on clinical care pathways. We present, here, the development of an HF DataMart framework for the management of clinical and research processes.MethodsWithin our institution, Fondazione Policlinico Universitario A. Gemelli in Rome (Italy), a digital platform dedicated to HF patients has been envisioned (GENERATOR HF DataMart), based on two building blocks: 1. All retrospective information has been integrated into a multimodal, longitudinal data repository, providing in one single place the description of individual patients with drill-down functionalities in multiple dimensions. This functionality might allow investigators to dynamically filter subsets of patient populations characterized by demographic characteristics, biomarkers, comorbidities, and clinical events (e.g., re-hospitalization), enabling agile analyses of the outcomes by subsets of patients. 2. With respect to expected long-term health status and response to treatments, the use of the disease trajectory toolset and predictive models for the evolution of HF has been implemented. The methodological scaffolding has been constructed in respect of a set of the preferred standards recommended by the CODE-EHR framework.ResultsSeveral examples of GENERATOR HF DataMart utilization are presented as follows: to select a specific retrospective cohort of HF patients within a particular period, along with their clinical and laboratory data, to explore multiple associations between clinical and laboratory data, as well as to identify a potential cohort for enrollment in future studies; to create a multi-parametric predictive models of early re-hospitalization after discharge; to cluster patients according to their ejection fraction (EF) variation, investigating its potential impact on hospital admissions.ConclusionThe GENERATOR HF DataMart has been developed to exploit a large amount of data from patients with HF from our institution and generate evidence from real-world data. The two components of the HF platform might provide the infrastructural basis for a combined patient support program dedicated to continuous monitoring and remote care, assisting patients, caregivers, and healthcare professionals.

Abstract The COVID-19 pandemic is impressively challenging the healthcare system. Several prognostic models have been validated but few of them are implemented in daily practice. The objective of the study was to validate a machine-learning risk prediction model using easy-to-obtain parameters to help to identify patients with COVID-19 who are at higher risk of death. The training cohort included all patients admitted to Fondazione Policlinico Gemelli with COVID-19 from March 5, 2020, to November 5, 2020. Afterward, the model was tested on all patients admitted to the same hospital with COVID-19 from November 6, 2020, to February 5, 2021. The primary outcome was in-hospital case-fatality risk. The out-of-sample performance of the model was estimated from the training set in terms of Area under the Receiving Operator Curve (AUROC) and classification matrix statistics by averaging the results of fivefold cross validation repeated 3-times and comparing the results with those obtained on the test set. An explanation analysis of the model, based on the SHapley Additive exPlanations (SHAP), is also presented. To assess the subsequent time evolution, the change in paO2/FiO2 (P/F) at 48 h after the baseline measurement was plotted against its baseline value. Among the 921 patients included in the training cohort, 120 died (13%). Variables selected for the model were age, platelet count, SpO2, blood urea nitrogen (BUN), hemoglobin, C-reactive protein, neutrophil count, and sodium. The results of the fivefold cross-validation repeated 3-times gave AUROC of 0.87, and statistics of the classification matrix to the Youden index as follows: sensitivity 0.840, specificity 0.774, negative predictive value 0.971. Then, the model was tested on a new population (n = 1463) in which the case-fatality rate was 22.6%. The test model showed AUROC 0.818, sensitivity 0.813, specificity 0.650, negative predictive value 0.922. Considering the first quartile of the predicted risk score (low-risk score group), the case-fatality rate was 1.6%, 17.8% in the second and third quartile (high-risk score group) and 53.5% in the fourth quartile (very high-risk score group). The three risk score groups showed good discrimination for the P/F value at admission, and a positive correlation was found for the low-risk class to P/F at 48 h after admission (adjusted R-squared = 0.48). We developed a predictive model of death for people with SARS-CoV-2 infection by including only easy-to-obtain variables (abnormal blood count, BUN, C-reactive protein, sodium and lower SpO2). It demonstrated good accuracy and high power of discrimination. The simplicity of the model makes the risk prediction applicable for patients in the Emergency Department, or during hospitalization. Although it is reasonable to assume that the model is also applicable in not-hospitalized persons, only appropriate studies can assess the accuracy of the model also for persons at home.

PurposeDistant metastases are currently the main cause of treatment failure in locally advanced rectal cancer (LARC) patients. The aim of this research is to investigate a correlation between the variation of radiomics features using pre- and post-neoadjuvant chemoradiation (nCRT) magnetic resonance imaging (MRI) with 2 years distant metastasis (2yDM) rate in LARC patients.Methods and MaterialsDiagnostic pre- and post- nCRT MRI of LARC patients, treated in a single institution from May 2008 to June 2015 with an adequate follow-up time, were retrospectively collected. Gross tumor volumes (GTV) were contoured by an abdominal radiologist and blindly reviewed by a radiation oncologist expert in rectal cancer. The dataset was firstly randomly split into 90% training data, for features selection, and 10% testing data, for the validation. The final set of features after the selection was used to train 15 different classifiers using accuracy as target metric. The models’ performance was then assessed on the testing data and the best performing classifier was then selected, maximising the confusion matrix balanced accuracy (BA).ResultsData regarding 213 LARC patients (36% female, 64% male) were collected. Overall 2yDM was 17%. A total of 2,606 features extracted from the pre- and post- nCRT GTV were tested and 4 features were selected after features selection process. Among the 15 tested classifiers, logistic regression proved to be the best performing one with a testing set BA, sensitivity and specificity of 78.5%, 71.4% and 85.7%, respectively.ConclusionsThis study supports a possible role of delta radiomics in predicting following occurrence of distant metastasis. Further studies including a consistent external validation are needed to confirm these results and allows to translate radiomics model in clinical practice. Future integration with clinical and molecular data will be mandatory to fully personalized treatment and follow-up approaches.

Background: In recent years, novel radiation therapy techniques have moved clinical practice toward tailored medicine. An essential role is played by the decision support system, which requires a standardization of data collection. The Aim of the Prediction Models In Stereotactic External radiotherapy (PRE.M.I.S.E.) project is the implementation of systems that analyze heterogeneous datasets. This article presents the project design, focusing on brain stereotactic radiotherapy (SRT). Materials & methods: First, raw ontology was defined by exploiting semiformal languages (block and entity relationship diagrams) and the natural language; then, it was transposed in a Case Report Form, creating a storage system. Results: More than 130 brain SRT’s variables were selected. The dedicated software Beyond Ontology Awareness (BOA-Web) was set and data collection is ongoing. Conclusion: The PRE.M.I.S.E. project provides standardized data collection for a specific radiation therapy technique, such as SRT. Future aims are: including other centers and validating an extracranial SRT ontology.

Purpose : Aim of the COBRA (Consortium for Brachytherapy Data Analysis) project is to create a multicenter group (consortium) and a web-based system for standardized data collection. Material and methods: GEC-ESTRO (Groupe Européen de Curiethérapie – European Society for Radiotherapy & Oncology) Head and Neck (H&N) Working Group participated in the project and in the implementation of the consortium agreement, the ontology (data-set) and the necessary COBRA software services as well as the peer reviewing of the general anatomic site-specific COBRA protocol. The ontology was defined by a multicenter task-group. Results : Eleven centers from 6 countries signed an agreement and the consortium approved the ontology. We identified 3 tiers for the data set: Registry (epidemiology analysis), Procedures (prediction models and DSS), and Research (radiomics). The COBRA-Storage System (C-SS) is not time-consuming as, thanks to the use of “brokers”, data can be extracted directly from the single center’s storage systems through a connection with “structured query language database” (SQL-DB), Microsoft Access®, FileMaker Pro®, or Microsoft Excel®. The system is also structured to perform automatic archiving directly from the treatment planning system or afterloading machine. The architecture is based on the concept of “on-purpose data projection”. The C-SS architecture is privacy protecting because it will never make visible data that could identify an individual patient. This C-SS can also benefit from the so called “distributed learning” approaches, in which data never leave the collecting institution, while learning algorithms and proposed predictive models are commonly shared. Conclusions : Setting up a consortium is a feasible and practicable tool in the creation of an international and multi-system data sharing system. COBRA C-SS seems to be well accepted by all involved parties, primarily because it does not influence the center’s own data storing technologies, procedures, and habits. Furthermore, the method preserves the privacy of all patients.

The new era of cognitive health care systems offers a large amount of patient data that can be used to develop prediction models and clinical decision support systems. In this frame, the multi-institutional approach is strongly encouraged in order to reach more numerous samples for data mining and more reliable statistics. For these purposes, shared ontologies need to be developed for data management to ensure database semantic coherence in accordance with the various centers’ ethical and legal policies. Therefore, we propose a privacy-preserving distributed approach as a preliminary data analysis tool to identify possible compliance issues and heterogeneity from the agreed multi-institutional research protocol before training a clinical prediction model. This kind of preliminary analysis appeared fast and reliable and its results corresponded to those obtained using the traditional centralized approach. A real time interactive dashboard has also been presented to show analysis results and make the workflow swifter and easier.