Deep learning algorithms already support medical imaging in many areas today and include all methods, such as X-ray, ultrasound, CT and MRI examinations. The increasing automation of diagnostic procedures offers advantages for patients and physicians alike: they increase the accuracy and reliability of examination results, facilitate diagnostic statements and enable improved, individually tailored therapeutic measures.
The basic prerequisite for the use of Artificial Neural Networks (ANN) is in many cases machine vision. Depending on the application and requirements, three different vision system architectures can be outlined for ANNs today, which differ in terms of performance, engineering effort and Total Cost of Ownership (TCO), in some cases significantly:
- embedded systems
- PC-based systems
- FPGA frame grabber-based systems
In the White Paper, Basler Product Manager Peter Behringer explains the four steps of a typical image processing process and compares the three types of vision systems for deep learning.
Furthermore, the White Paper provides valuable insights on
- deep learning in microscopy,
- digital pathology,
- virtual staining in microscopy and on
- how to make artificial neural network-based products commercially viable.
