Clinical Application Value of Deep Learning-Based "Opportunistic Screening" for Malignant Tumors on Routine Non-Contrast Chest-Abdomen-Pelvis CT

Clinical Application Value of Deep Learning-Based "Opportunistic Screening" for Malignant Tumors on Routine Non-Contrast Chest-Abdomen-Pelvis CT

This study aims to develop and validate a deep learning-based opportunistic multi-cancer screening system using routine non-contrast chest-abdomen-pelvis CT examinations, including CHANCE-Breast, CHANCE-Liver, CHANCE-Kidney, and CHANCE-Bladder, for the early detection of breast, liver, kidney, and bladder cancers. In addition, the study will assess a human-AI collaborative framework to determine its potential for improving cancer detection and reducing missed diagnoses in clinical practice.

Inclusion Criteria:

1. Patients with a confirmed diagnosis of the target malignancy who received treatment at our institution;
2. Diagnostic-quality CT images without substantial metal or motion artifacts and with complete anatomical coverage of the target organ (breast, liver, kidney, or bladder);
3. Availability of complete pre-treatment non-contrast CT imaging data.

Exclusion Criteria:

1. Non-diagnostic image quality;
2. Absence of a definitive reference-standard diagnosis;
3. Incomplete clinical or imaging data.