Period of realization: 2022 – 2024
Plan of research with objectives
The research activities in this project will belong to one of the following groups
The sample preparation and data acquisition:
● Small laboratory animals, habituation and separation into experimental groups;
● Study of the radiobiological effects of various types of ionizing radiation with different physical characters and medical agents on the central nervous system;
● Biopsy sample collection, handling, and processing for the subsequent pathologic evaluation, together with a gathering of complete specimen information, and preparation of samples for the transport;
● Optimization of the sample fixation and staining techniques to facilitate the subsequent automatic image segmentation and signal processing by the information system. The overall goal is to increase the color contrast between relevant parts of the sample and the remaining background;
● Classification and systematization of the obtained data for inclusion into the developed database and further investigations.
Data processing and analysis:
● Development of algorithms for processing experimental data based on automated and deep learning methods for pathomorphological and behavioral analysis;
● Development of an automated information system (AIT) based on modern IT web technologies providing interconnection of multiple platforms and data sharing, management of the acquired data, its permanent storage, automated standard statistical analysis, and visualization of obtained data;
● Development of specialized algorithms based on deep learning models, modern inference solutions, for automated data segmentation (i.e., separation of the neural cells or collagen fibers from the background), and characterization (i.e., determination of the number and size of neural cells or collagen density, porosity, etc.).
Modeling of the obtained data
● Mathematical formalization of biologically relevant concepts, and their quantification (For example, how to capture and quantify the straightness of collagen fiber distribution, its thickness, porosity, etc.), thus enabling concise description of the investigated sample by a list of the relevant biological parameters;
● Optimization of the developed models to be suitable for implementation in artificial neural networks;
● Development of the methods for systematic investigation of the appropriate subspaces of the parametric space for the presence of catastrophic flags such as modality, sudden jumps, hysteresis, divergences, and anomalous dispersion. which are signatures of certain prototypical dependence between parameters of the system known from the singularity and catastrophe theories.
The progress of human civilization has brought a tremendous increase of well-being to society with the hidden cost paid by the unintended side-effects affecting the health of individuals. This is especially evident in our late stage of development where we are constantly exposed to various combinations of chemicals and the presence of ionizing and non-ionizing radiation. Therefore, it is not surprising that our societies are faced with the increased numbers of malignant diseases.
In addition, the reduction of the carbon footprint of our civilization on mother Earth will probably demand increased reliance on nuclear energy. The conditions of exposure to humans and the forms of radiation damage that are formed at the same time become more diverse. This includes flights outside the magnetosphere. where the absence of real experience with long space flights outside the near-Earth orbit prevents assessment of the possible risks from exposure to ionizing radiation. All those tasks require heavy reliance on models of human reaction to the presence of the various cancerogenic agents, which is possible only on experimental animals, which allows more efficient evaluation of the possible consequences of the functioning of the organs.
The developed software components, their testing, approbation, and final implementation of specialized AIS will be carried out on the basis of the already existing Multifunctional Information and Computing Complex of LIT JINR, which will facilitate the smooth introduction of the new technology. Its implementation will make it possible to conduct a comprehensive analysis of heterogeneous experimental data, as well as from various research groups, and to almost fully automate data analysis and presentation of results, which, in conclusion, will accelerate the production of qualitatively new results. The developed service can be used for medical, research, and educational purposes. The suggested use of machine learning methods enables continuous perfection of developed models by new data. Such a system can be used to train new pathology residencies.
The structural stability of designed models guarantees its applicability to the much wider area of the parametric space used for its construction. Phenomenological models constructed mainly from experimental observations provide a causal relationship between system parameters, while topology and singularity theory enable simple identification of dominant parameters driving observed morphological changes of the examined samples and their biological function. This knowledge combined with the developed AIS could indicate new ways for diagnostic of the early development of cancer, and its prevention. If it turns possible to set values of the parameters by the exogenous means then developed models will be useful in the development of the new cancer treatments.