CVPR Daily - Wednesday

DAILY Wednesday Medical Computer Vision 15 Lena Maier-Hein from DFKZ described some exciting work on handling uncertainties in surgical image processing in a domain coined: “surgical data science” . She described how building trust in machine learning - based solutions is an important prerequisite for future clinical translation. She described methods for handling ambiguities in inverse problems and out-of-distribution detection in spectral image analysis. Purang Abolmaesumi , University of British Columbia, showcased exciting work on integrating and evaluating various AI technologies for echocardiography imaging in orde ecision making. Georg Langs , Medical University of Vienna, gave an exciting talk on developing machine learning models for discovering medical imaging markers for disease progression, with the objective of using them to make better treatment decisions. Qi Dou , Chinese University of Hong Kong, described new methods for improving deep model generalization for medical images, including: supervised multi-site learning, unsupervised domain adaptation, meta-learning based domain generalization. Julia Schnabel , King’s College London, gave an exciting presentation on using deep learning for "smart" cardiac magnetic resonance imaging, with an emphasis on improving image quality, and on integrated solutions for obtaining high-quality segmentations from raw motion-corrupted or poor-quality sensor data.