Figure 8 - uploaded by Andreas Maier
Content may be subject to copyright.
Source publication
CT is doubtlessly one of the most important technologies in medical imaging and offers us views inside the human body that are as valuable to physicians as they are fascinating (cf. Fig. 8.1).
Citations
... This comes from the need of sampling correctly in the Fourier space, based on the central slice theorem. This is a the fundamental concept of the tomography theory and reconstruction, and is extensively covered in the literature, for example in [56]. ...
X-ray ptychography, a scanning coherent diffraction imaging technique, is one of the most used techniques at synchrotron facilities for high resolution imaging, with applications spanning from life science to nano-electronics. In the recent years there has been a great effort to make the technique faster to enable high throughput nanoscale imaging. Here we apply a fast ptychography scanning method to image in 3D $$10^6\upmu \mathrm{{m}}^3$$ 10 6 μ m 3 of brain-like phantom at 3 kHz, in a 7 h acquisition with a resolution of 270 nm. We then present the latest advances in fast ptychography by showing 2D images acquired at 110 kHz by combining the fast-acquisition scheme with a high-acquisition rate prototype detector from DECTRIS Ltd. We finally review the experimental outcome and discuss the prospective use of fast ptychography schemes for the investigation of mm $$^{3}$$ 3 size samples of brain-like phantom, by extrapolating the current results to the high coherent flux scenario of diffraction limited storage rings.
Reliably detecting diseases using relevant biological information is crucial for real-world applicability of deep learning techniques in medical imaging. We debias deep learning models during training against unknown bias – without preprocessing/filtering the input beforehand or assuming specific knowledge about its distribution or precise nature in the dataset. Control regions are used as surrogates that carry information regarding the bias; subsequently, the classifier model extracts features, and biased intermediate features are suppressed by our custom, modular DecorreLayer. We evaluate our method on a dataset of 952 lung computed tomography scans by introducing simulated biases w.r.t. reconstruction kernel and noise level and propose including an adversarial test set in evaluations of bias reduction techniques. In a moderately sized model architecture, applying the proposed method to learn from data exhibiting a strong bias, it near-perfectly recovers the classification performance observed when training with corresponding unbiased data.
Studies have shown that artificial intelligence is a novel realm of science and technology rapidly expanding globally across many industrial sectors. The recent advances in the artificial intelligence system in medicine and healthcare systems, particularly neuroscience and clinical practice, have presented extraordinary opportunities such as disease prediction, communication, and treatment strategies. There are many genetic tests, existing software, personal monitoring devices, editing tools, online platforms, personalized digital models, companion and surgical robotics, and augmented reality devices adopted in medical practice to monitor and analyze cognitive therapy. Artificial intelligence has been utilized to implement and analyze medical images, videos, nodules, lesions, and signals. Thus, this chapter will give a detailed account of different artificial intelligence models for risk assessment, classification, segmentation tasks, prognosis, diagnosis, and prediction of therapy reaction in cognitive therapy.
