Aiding radiological thorax CT assessments in UZ Leuven using the contextflow AI-based algorithms

The contextflow tool is CE-marked (Class IIa) medical device software that uses deep learning models to support radiologists in interpreting chest CT images.

Core Features: The software provides computer-aided detection for lung cancer, ILD, and IPE. Its key functionalities include automated nodule detection, quantification, and tracking; a malignancy risk score (MSI); quantitative analysis of 19 lung tissue patterns; and detection of incidental findings like pulmonary embolism.

Functionalities:

• Lung Pattern Detection (for ILDs): This feature provides advanced computer-aided support for complex cases like Interstitial Lung Disease (ILD) by automatically performing a quantitative and qualitative analysis of 19 different lung tissue image patterns. This helps in the detection and characterization of diffuse lung diseases.
• Lung Nodule Detection: The software automatically scans chest CT images to detect and quantify suspected lung nodules. This core function is designed to assist in the early detection of potential lung cancers.
• Nodule TIMELINE Tracking: For follow-up examinations, this tool automatically recognizes and tracks nodules over time by comparing current scans to previous ones. It provides valuable data on changes in nodule volume and offers RECIST scores to monitor growth or response to treatment.
• Nodule Malignancy Prediction (mSi): The Malignancy Similarity Index (MSI) is an AI-driven feature that calculates a probability score for a detected lung nodule, indicating its likelihood of being benign or malignant. This helps radiologists risk-stratify nodules and may reduce unnecessary biopsies.
• Incidental Pulmonary Embolism (IPE) Detection: Acting as a safety net, this functionality is designed to detect incidental pulmonary embolisms on chest CTs that were ordered for other clinical reasons.
• Similar Case SEARCH: This unique decision-support tool functions like a “Google for lung patterns,” allowing radiologists to search a large database of pre-labelled cases for visually similar patterns. This provides valuable reference cases and differential diagnosis information to aid in complex interpretations.
• As of November 2025 contextflow will add Coronary Artery Calcification (CAC) detection and risk assessment.