Taking AI to the next level with Data Centric approach in Computational Pathology
In the field of AI and in AI for Digital Pathology in particular, the focus is usually on model-centric approaches, e.g improved solutions are created by skilled researchers and engineers that improve the models, with a fixed dataset. However lately the “data-centric” approach is gaining attention, where the models are fixed and the focus is on improving the dataset. In a recent perspective presented by Andrew NG, to achieve a good AI solution there should be a balance between a model-centric vs a data-centric approaches, and having a high quality data is essential. In this lecture we will discuss model centric vs data centric approaches. We will show how data-centric approaches are especially suitable for Digital Pathology. We will go over real examples of how we aim to achieve this with concepts like Rapid data collection, Active Learning, Interactive Learning, and Identifying Mislabeled annotations with the DeePathology STUDIO.
Data-efficient and multimodal computational pathology
Advances in digital pathology and artificial intelligence have presented the potential to build assistive tools for objective diagnosis, prognosis and therapeutic-response and resistance prediction. In this talk we will discuss: 1) Data-efficient methods for weakly-supervised whole slide classification with examples in cancer diagnosis and subtyping, allograft rejection etc. (Nature Biomedical Engineering, 2021). 2) Harnessing weakly-supervised, fast and data-efficient WSI classification for identifying origins for cancers of unknown primary (Nature, 2021). 3) Discovering integrative histology-genomic prognostic markers via interpretable multimodal deep learning (IEEE TMI, 2020). 4) Deploying weakly supervised models in low resource settings without slide scanners, network connections, computational resources and expensive microscopes. 5) Bias and fairness in computational pathology algorithms.
AI in pathology: promise and potential
The emerging discipline of AI-based computational pathology has shown great promise. Integration of informatics along with good computing capacity has helped revolutionize AI-based computational pathology The field of computational pathology has grown beyond counting nuclei/mitosis or classifying tumour tissue, and is increasingly incorporating the complex process of analysis and judgment using demographic information, digital pathology, -omics, and laboratory results. Deep learning has been previously applied to automated biomarker assessment, and now aims to learn biomarkers from HE, which could facilitate additional tools for pathologists, and transform the field of companion diagnostics and drug discovery. Despite the technical, ethical, and regulatory challenges, computational pathology is a valuable synergistic system, which can provide insights into the diagnosis, prognosis, and treatment of disease. Computational pathology has the potential to improve diagnostic accuracy, thereby optimizing patient care.