Innovative AI System Enhances Liver Tumor Detection and Monitoring

Researchers at the Vall d'Hebron Institute of Oncology have introduced SALSA, an advanced artificial intelligence tool designed to revolutionize the detection and tracking of liver tumors, including primary hepatocellular carcinoma and metastatic lesions. Developed by the VHIO Radiomics Group under the leadership of Raquel Perez-Lopez, SALSA employs deep learning algorithms to analyze medical images—primarily computed tomography (CT) scans—with remarkable precision. This fully automated system accurately identifies and delineates liver tumors, reducing the variability and time-consuming efforts traditionally associated with manual contouring by radiologists.

The development process involved training SALSA on a comprehensive dataset comprising 1598 CT scans and over 4908 tumor lesions, utilizing the nnU-Net segmentation approach. The results demonstrated that SALSA surpasses current state-of-the-art models, achieving a tumor detection accuracy exceeding 99% per patient and nearly 82% lesion-by-lesion detection in external validation cohorts. These findings suggest that SALSA can significantly improve diagnostic accuracy, assist in treatment planning, and provide reliable tumor burden quantification—crucial factors for prognosis and therapeutic decisions.

Imaging plays a vital role in oncology, offering essential insights for cancer detection, staging, and response assessment. However, manual tumor delineation poses challenges due to its complexity and variability among observers, often creating bottlenecks in research and clinical workflows. By automating these processes, SALSA addresses these issues, streamlining cancer evaluation.

Beyond detection, SALSA’s capabilities extend to aiding the discovery of imaging biomarkers such as tumor volume, density, and texture—valuable parameters for evaluating treatment response. Given that current clinical criteria for assessing tumor response are limited—often based solely on 2D measurements—SALSA’s comprehensive volumetric analysis can support personalized treatment approaches and improve outcomes for patients with liver cancer.

The validation across diverse patient groups confirms SALSA’s reliability and surpasses radiologists’ accuracy, making it a promising tool in clinical oncology. This innovation highlights the significant potential of AI-driven solutions to transform cancer management by facilitating precise, automated, and efficient tumor analysis.

For more details, refer to the original publication: DOI: 10.1016/j.xcrm.2025.102032. Source: https://medicalxpress.com/news/2025-05-ai-tool-automates-liver-tumor.html