AI Tool SALSA Automates Precise Detection and Monitoring of Liver Tumors, Outperforming Radiologists

Researchers from the Vall d'Hebron Institute of Oncology's (VHIO) Radiomics Group, led by Raquel Perez-Lopez, have developed a new AI tool called SALSA (System for Automatic Liver tumor Segmentation And detection). This innovation, detailed in a recent publication in Cell Reports Medicine, aims to automate the detection and monitoring of liver tumors, specifically hepatocellular carcinoma (HCC) and metastatic tumors. Medical imaging, especially CT scans, plays a critical role in cancer diagnosis, treatment planning, and response evaluation. However, the process of precisely delineating tumors for volumetric analysis can be both time-consuming and subject to variability among different observers. This bottleneck can delay research projects and clinical applications, making accurate and consistent tumor contouring a significant challenge. To address these issues, the VHIO team developed SALSA, which utilizes deep learning techniques to detect and segment liver tumors automatically and accurately. The tool was trained using the existing nnU-Net segmentation method and data from 1,598 CT scans featuring 4,908 primary or metastatic liver tumors. The outcome is a system that demonstrates superior accuracy for cancer detection and precise quantification of tumor burden, outperforming both state-of-the-art models and the inter-agreement levels among radiologists. Maria Balaguer-Montero, a Ph.D. student and first author of the study, reported that SALSA achieved a precision rate of over 99% at the patient level in detecting liver tumors and nearly 82% in lesion-by-lesion detection during external validation. These results highlight SALSA's effectiveness and reliability across various test cohorts, often matching and even exceeding the performance of expert radiologists. The ability to automatically and precisely identify liver tumors has several important implications. First, it can enhance the accuracy and efficiency of tumor detection, thereby improving early diagnosis and the overall treatment planning process. Second, it enables a more comprehensive assessment of tumor burden, which is crucial for predicting cancer progression and determining the effectiveness of treatments. Current clinical response criteria are somewhat limited, relying on 2D measurements of tumor diameter and evaluating only two tumors per organ and up to five lesions per patient. SALSA, on the other hand, can measure parameters like total tumor volume, density, and texture, providing a more holistic view of treatment response. This capability could significantly support therapeutic decision-making and lead to better patient outcomes. Perez-Lopez emphasized that the tool could be particularly beneficial in managing liver cancer cases, where precise tumor quantification is essential for evaluating the response to therapy. By enabling the measurement of multiple tumors and their characteristics, SALSA can offer a more detailed and accurate assessment, thereby enhancing the precision of personalized oncology. The Radiomics Group at VHIO is at the forefront of developing medical imaging tools that leverage imaging biomarkers, such as tumor volume, density, and texture, to improve cancer treatment. The group's work aligns with the broader trend of using AI and machine learning to advance personalized medicine. This approach not only streamlines the diagnostic and monitoring processes but also holds the potential to customize treatment plans based on the unique characteristics of each patient's tumors. In addition to its immediate clinical applications, SALSA could also have significant research implications. It can accelerate the analysis of large datasets, reducing the time and resources required for manual tumor segmentation, and facilitate more robust and reliable studies in oncology. Industry insiders and experts applaud the development of SALSA, noting its potential to revolutionize the way liver tumors are managed. The tool not only improves the precision and consistency of tumor detection but also supports a more data-driven approach to oncology, which could lead to more effective and personalized treatments. The Vall d'Hebron Institute of Oncology, known for its cutting-edge research and clinical trials, continues to push the boundaries of cancer care through innovative technologies like SALSA.