LAICSIDE
![LAICSIDE image](https://aiforoncology.nl/media/project_photos/Designer-3.jpeg)
2024-2028
The digitization of tumor tissue slides enables automated computational pathology analysis, offering more objective pattern recognition and insights into patient data. Digital workflows and AI-powered tools enable pathologists to improve diagnostic precision, reproducibility, and scalability. In addition to tailoring patient treatment and enhancing outcomes, it reduces pathologists' workload hence addressing the shortage of specialized staff in the face of an ever-increasing cancer burden.
Despite progress, many challenges persist. Computational pathology often faces dataset issues and the related AI solutions can often lack interpretability, impacting safety and clinical acceptance. Moreover, recent developments in cancer research have shown the importance of investigating cell spatial organizations, including their potential positive impact on interpretability, which until today has often been overlooked.
LAICSIDE’s objectives address these challenges by exploring cell-level segmentation and tissue spatial organization through training a cell phenotyper (CELINE) on a large and diverse annotated single-cell dataset and integrating it with a large multimodal model (coPATHY). CELINE aims to identify any cell and cell type in H&E-stained WSIs, a novelty compared to narrow single-cell-type segmentation models. The introduction of coPATHY yields an interactive platform facilitating multimodal (text and image) slide exploration, interpretability, and diagnosis capabilities, a significant advance in digital and AI-driven pathology.
Offering a comprehensive pan-cancer cell annotation dataset and novel approaches to workflow automation, we aim to revolutionize pathology through AI. By understanding tumor cell spatial complexity, CELINE and coPATHY enable precise cell-level analysis critical for classifying cancer subtypes. Anticipated to streamline diagnostic processes, the project promises more accurate and efficient cancer diagnosis, enhancing precision pathology and patient care in breast and prostate cancer and beyond.
This project is a partnership with kaiko.ai and is funded by Health Holland.