In this project we aim to develop and validate AI techniques for the detection of serous tubal intra-epithelial carcinoma (STIC), a non-invasive lesion in the distal Fallopian tube which is expected to be a precursor for ovarian cancer.
Maschenka Balkenhol succesfully defended her PhD thesis titled 'Tissue-based biomarker assessment for predicting prognosis of triple negative breast cancer: the additional value of artificial intelligence' on the 15th of September.
The impact of scanner variations and stain normalization on CNN performance for prostate cancer classification on WSIs was investigated by Zaneta Swiderska-Chadaj and their colleagues, and the work was published in Nature Scientific Reports.
Due to memory constraints on current hardware, most convolutional neural networks (CNN) are trained on sub-megapixel images. A novel method for end-to-end training of CNNs on multi-megapixel images was proposed by Hans Pinckaers and his colleagues. Their work appeared online in IEEE Transactions on Pattern Analysis and Machine Intelligence.
RSNA News interviewed Keelin Murphy about her Radiology publication on CAD4COVID-XRay, an AI solution for finding COVID-19 with chest radiography.
Together with colleagues from the NCI and the AMC, DIAG's Jonas Teuwen and Nikita Moriakov won an international competition where MRI-scans can be accelerated using deep learning.
Pulmonary lobe segmentation in computed tomography scans is essential for regional assessment of pulmonary diseases. Our algorithm for automatic segmentation of pulmonary lobes on CT scans for patients with COPD or COVID-19 is now available on Grand Challenge.
After CAMELYON, there is now PANDA, our new challenge on prostate cancer grading in collaboration with Kaggle and Karolinska Institutet. Our aim is to crowdsource the best possible solution to help pathologists better diagnose and treat patients.
The Novartis Transplantation Award for Basic Research was awarded to the Computational Pathology group for the JASN publication on Deep learning for histopathologic renal tissue assessment by Meyke Hermsen et al.
The Gleason score suffers from significant inter-observer variability. This problem could be solved by the fully automated deep learning system developed by Wouter Bulten and his colleagues. Their work appeared online today in The Lancet Oncology.