CLOBNET

CLOBNET enables integration and machine learning analysis of structured electronic health record (EHR) data to support prediction of therapy outcomes in high-grade serous ovarian cancer (HGSOC).

Key Features:

• ETL Capabilities: Performs extract, transform, and load (ETL) processing to integrate and transform complex healthcare datasets into analyzable formats.
• Structured EHR Data Processing: Processes and supports analysis and visualization of structured EHR data for downstream modeling.
• Machine Learning Models: Employs machine learning models, including logistic regression, for outcome prediction and analysis of clinical data.
• Dissemination Score: Implements a novel dissemination score that quantifies disease burden at diagnosis.
• Outcome Definition: Uses RECIST criteria (version 1.1) to define outcomes such as progressive disease (PD) and complete response (CR).
• Model Evaluation Metrics: Reports performance metrics including area under the ROC curve (AUROC 0.86) and specificity (73%) and sensitivity (89%) for the best-performing logistic regression model.
• Cohort Integration: Integrates comprehensive clinical and EHR data from a prospective HGSOC cohort of 208 patients treated at Turku University Hospital between 2009 and 2019.

Scientific Applications:

• Therapy Outcome Prediction in HGSOC: Predicts primary therapy outcomes (PD versus CR) in high-grade serous ovarian cancer using structured EHR and clinical data.
• Quantification of Disease Burden: Applies a dissemination score to quantify disease burden at diagnosis for use in predictive modeling.
• Model Validation on Clinical Cohort: Demonstrates model performance on a prospective cohort of 208 patients from Turku University Hospital (2009–2019).

Methodology:

Performs ETL on structured EHR data, computes a dissemination score, trains machine learning models (including logistic regression), and evaluates performance using AUROC, sensitivity, and specificity with outcomes defined by RECIST criteria (version 1.1).