Date(s) - 17/07/2017
11:00 am - 12:00 pm
Studio Villa Bosch
Machine Learning Based Image Analysis for Supporting Microscopic Tissue Analysis in Cancer
Despite sequencing technology generating large amounts of patient individual data, the actual diagnostics and definition of major diseases is done through pathology. But this microscopic analysis of tissue sections and cells is still refraining from digitization and digital analysis. In consequence, critical treatment decisions remain subjective and depending on high quality pathology readers. Moreover, our general understanding of diseases, drug development and clinical trials is currently lacking a digital data basis. Through its ability of learning highly complex patterns in pre-defined network structures, deep learning is a pre-destined tool for medical image analysis. In the talk, some applications and lessons-learned from applying machine and deep learning for microscopic tissue analysis are discussed with an outlook on tissue simulation. In diagnostics, from a collaboration with the US National Cancer Institute (NCI), an independently validated deep learning based assay (CINREADER) for screening for cervical cancer from cytological specimen is shown, being slightly superior to human expert reading. As an assistant, CINREADER could increase screening expert performance significantly and thus help distributing diagnostic standards globally. The case study illustrates the potential of machine learning in improving cancer care, but also pinpoints its challenges.
Curriculum vitae: cv_ng_hits
For registration please contact Benedicta Frech: firstname.lastname@example.org