Handling of uncertainty in medical data using machine learning and probability theory techniques: a review of 30 years (1991-2020).

Understanding the data and reaching accurate conclusions are of paramount importance in the present era of big data. Machine learning and probability theory methods have been widely used for this purpose in various fields. One critically important yet less explored aspect is capturing and analyzing uncertainties in the data and model. Proper quantification of uncertainty helps to provide valuable information to obtain accurate diagnosis. This paper reviewed related studies conducted in the last 30 years (from 1991 to 2020) in handling uncertainties in medical data using probability theory and machine learning techniques. Medical data is more prone to uncertainty due to the presence of noise in the data. So, it is very important to have clean medical data without any noise to get accurate diagnosis. The sources of noise in the medical data need to be known to address this issue. Based on the medical data obtained by the physician, diagnosis of disease, and treatment plan are prescribed. Hence, the uncertainty is growing in healthcare and there is limited knowledge to address these problems. Our findings indicate that there are few challenges to be addressed in handling the uncertainty in medical raw data and new models. In this work, we have summarized various methods employed to overcome this problem. Nowadays, various novel deep learning techniques have been proposed to deal with such uncertainties and improve the performance in decision making.

Hybrid multiscale modeling and prediction of cancer cell behavior.

BACKGROUND: Understanding cancer development crossing several spatial-temporal scales is of great practical significance to better understand and treat cancers. It is difficult to tackle this challenge with pure biological means. Moreover, hybrid modeling techniques have been proposed that combine the advantages of the continuum and the discrete methods to model multiscale problems. METHODS: In light of these problems, we have proposed a new hybrid vascular model to facilitate the multiscale modeling and simulation of cancer development with respect to the agent-based, cellular automata and machine learning methods. The purpose of this simulation is to create a dataset that can be used for prediction of cell phenotypes. By using a proposed Q-learning based on SVR-NSGA-II method, the cells have the capability to predict their phenotypes autonomously that is, to act on its own without external direction in response to situations it encounters. RESULTS: Computational simulations of the model were performed in order to analyze its performance. The most striking feature of our results is that each cell can select its phenotype at each time step according to its condition. We provide evidence that the prediction of cell phenotypes is reliable. CONCLUSION: Our proposed model, which we term a hybrid multiscale modeling of cancer cell behavior, has the potential to combine the best features of both continuum and discrete models. The in silico results indicate that the 3D model can represent key features of cancer growth, angiogenesis, and its related micro-environment and show that the findings are in good agreement with biological tumor behavior. To the best of our knowledge, this paper is the first hybrid vascular multiscale modeling of cancer cell behavior that has the capability to predict cell phenotypes individually by a self-generated dataset.

Integrating evolutionary game theory into an agent-based model of ductal carcinoma in situ: Role of gap junctions in cancer progression.

BACKGROUND AND OBJECTIVE: There are many cells with various phenotypic behaviors in cancer interacting with each other. For example, an apoptotic cell may induce apoptosis in adjacent cells. A living cell can also protect cells from undergoing apoptosis and necrosis. These survival and death signals are propagated through interaction pathways between adjacent cells called gap junctions. The function of these signals depends on the cellular context of the cell receiving them. For instance, a receiver cell experiencing a low level of oxygen may interpret a received survival signal as an apoptosis signal. In this study, we examine the effect of these signals on tumor growth. METHODS: We make an evolutionary game theory component in order to model the signal propagation through gap junctions. The game payoffs are defined as a function of cellular context. Then, the game theory component is integrated into an agent-based model of tumor growth. After that, the integrated model is applied to ductal carcinoma in situ, a type of early stage breast cancer. Different scenarios are explored to observe the impact of the gap junction communication and parameters of the game theory component on cancer progression. We compare these scenarios by using the Wilcoxon signed-rank test. RESULTS: The Wilcoxon signed-rank test succeeds in proving a significant difference between the tumor growth of the model before and after considering the gap junction communication. The Wilcoxon signed-rank test also proves that the tumor growth significantly depends on the oxygen threshold of turning survival signals into apoptosis. CONCLUSIONS: In this study, the gap junction communication is modeled by using evolutionary game theory to illustrate its role at early stage cancers such as ductal carcinoma in situ. This work indicates that the gap junction communication and the oxygen threshold of turning survival signals into apoptosis can notably affect cancer progression.

Protein fold recognition with a two-layer method based on SVM-SA, WP-NN and C4.5 (TLM-SNC).

The structural knowledge of protein is crucial in understanding its biological role. An effort is made to assign a fold to a given protein in a protein fold recognition problem. A computational Two-Layer Method (TLM) based on the Support Vector Machine (SVM), the Neural Network (NN) and the Decision Tree (C4.5) has been developed in this study for the assignment of a protein sequence to a folding class in SCOP. Prediction accuracy is measured on a dataset and the accuracy of the proposed method is very promising in comparison with other classification methods.

Hepatitis disease diagnosis using a novel hybrid method based on support vector machine and simulated annealing (SVM-SA).

In this study, diagnosis of hepatitis disease, which is a very common and important disease, is conducted with a machine learning method. We have proposed a novel machine learning method that hybridizes support vector machine (SVM) and simulated annealing (SA). Simulated annealing is a stochastic method currently in wide use for difficult optimization problems. Intensively explored support vector machine due to its several unique advantages is successfully verified as a predicting method in recent years. We take the dataset used in our study from the UCI machine learning database. The classification accuracy is obtained via 10-fold cross validation. The obtained classification accuracy of our method is 96.25% and it is very promising with regard to the other classification methods in the literature for this problem.