Dr
Łukasz
Roszkowiak
Nałęcz Institute of Biocybernetics and Biomedical Engineering PAS, Warsaw
The application of deep learning is now a common practice. Convolutional neural networks took computer vision by storm. But is it really a long awaited panacea? In contrast to the classical approach, neural network models select the optimal set of features needed for classification. The bottleneck of these methods is the high demand for the labelled datasets needed for learning and generalisation of knowledge.
In medical domain the developed methods can support conducting research or assist in the diagnosis-making process. The main advantage is the objectivization of quantitative evaluation by increasing repeatability and reducing time of analysis. Fast and reliable diagnosis, more insightful prognosis or new understanding of disease mechanism could be established with the help of computer aided evaluation of digital images.
I have developed a new method for computer aided quantification of cells in whole-slide images of immunohistochemically stained breast cancer tissue samples. An implemented system named CHISEL (Computer-assisted Histopathological Image Segmentation and Evaluation) supports quantitative evaluation of cells in microscopic images. It is an end-to-end solution which is capable of processing benign and malignant tissue samples with nuclear staining of various intensity and diverse cellularity. The multi-algorithmic processing scheme that I used and the parameterization of the algorithms enable it to be quickly and efficiently adapted to different imaging data.
This is a hybrid event:
Room D, the Institute of Physics PAS, Al. Lotników 32/46
Online: Zoom Link, (Passcode: 134595, Meeting ID: 823 8038 0442)
