Virtual Lab
Machine Learning is involved! Our approach is based on image segmentation. It is established that image segmentation is the process of separating a digital image into multiple sets of pixels (also called segments or image objects) and is a prerequisite step to further image analysis to locate specific objects of interest. Our machine learning algorithms is able to: - Labelling pixels and identifying regions of interest (ROIs), - Provide objects of interest find in the image, - Automatically analyzing full datasets and specifying objects detected. It is undeniable that using automated image analysis to identify structures of interest makes the process exponentially faster and more efficient.
One of the artificial intelligence techniques used to analyze images (Virtual Slide), is the unsupervised learning. Let us recall that an unsupervised machine learning is a branch of machine learning, characterized by the analysis and grouping of unseated data. In particular, we have developed clustering alogorithms which are capable of detecting, segmenting and classifying images in the context of parasitological diagnosis of malaria. It should be noted that from a given input image, our algorithms do not know the exact output in advance. In practice the result of the unsupervised learning algorithm seems to be less accurate as input data is not labeled.
We did our experiment on Plasmodium Falciparum. The blood-stage parasites of Plasmodium Falciparum specie exhibit different morphologies. These blood-stages constitute classes or clusters of reference.
We have simulated 40 sample images from several sources, but here the majority is from the Division of Parasitic Diseases and Malaria (DPDM) of CDC. Below, we take up an example of image analysis proceeded by our VLAB system.
Evaluating the performance of a clustering algorithm is not as trivial as counting the number of errors or the precision and recall like in the case of supervised learning algorithms. However, there is a certain way to validate the results of our unsupervised learning model. The statistical test such as Kolmogorov-Smirnov(KS) can allow the validation. The test answers the question "How likely is it that we would see two sets of samples like this if they were drawn from the same (but unknown) probability distribution?".
Bergson and the holographic theory of mind Bergson’s model of time (1889) is perhaps the proto-phenomenological theory. It is part of a larger model of mind (1896) which can be seen in modern light as describing the brain as supporting a modulated wave within a holographic field, specifying the external image of the world, and wherein subject and object are differentiated not in terms of space, but of time. Bergson’s very concrete model is developed and deepened with Gibson’s ecological model of perception. It is applied to the problems of consciousness, direct realism, qualia and illusions. The model implies an entirely different basis for memory and cognition, and a brief overview is given for the basis of direct memory, compositionality and systematicity. Source: Bergson and the holographic theory of mind In Phenomenology and the Cognitive Sciences, 2006.
JPEG Pleno Holography Conventional imaging solutions adhere to a ray-based model to describe light propagation. In recent years, the popularity of holographic imaging solutions that follow an interference-based imaging model has increased significantly. Holographic microscopy and tomography provide a very high depth resolution and hence are extremely relevant in biomedical and industrial imaging and measurement. Moreover, n the creative media domain, AR/VR glasses and holographic table-top 6DoF displays are envisaged that utilized holographic technologies. The JPEG Pleno Holography efforts aims at providing solutions for the compression of this data that address - but not limited to - these above use cases. The JPEG committee invites the actors in this field to contribute and help to address the compliance issue existing in this field. JPEG Pleno is a standardization framework addressing the compression and signaling of plenoptic modalities. While the standardization of solutions to handle light field content is currently reaching its final stage, the Joint Photographic Experts Group (JPEG) committee is now preparing for the standardization of solutions targeting point cloud and holographic modalities Source: JPEG Pleno Holography. JPEG Pleno holography: scope and technology validation procedures.
