Application of explainable artificial intelligence for healthcare: A systematic review of the last decade (2011-2022).

BACKGROUND AND OBJECTIVES: Artificial intelligence (AI) has branched out to various applications in healthcare, such as health services management, predictive medicine, clinical decision-making, and patient data and diagnostics. Although AI models have achieved human-like performance, their use is still limited because they are seen as a black box. This lack of trust remains the main reason for their low use in practice, especially in healthcare. Hence, explainable artificial intelligence (XAI) has been introduced as a technique that can provide confidence in the model's prediction by explaining how the prediction is derived, thereby encouraging the use of AI systems in healthcare. The primary goal of this review is to provide areas of healthcare that require more attention from the XAI research community. METHODS: Multiple journal databases were thoroughly searched using PRISMA guidelines 2020. Studies that do not appear in Q1 journals, which are highly credible, were excluded. RESULTS: In this review, we surveyed 99 Q1 articles covering the following XAI techniques: SHAP, LIME, GradCAM, LRP, Fuzzy classifier, EBM, CBR, rule-based systems, and others. CONCLUSION: We discovered that detecting abnormalities in 1D biosignals and identifying key text in clinical notes are areas that require more attention from the XAI research community. We hope this is review will encourage the development of a holistic cloud system for a smart city.

Application of nonlinear methods to discriminate fractionated electrograms in paroxysmal versus persistent atrial fibrillation.

BACKGROUND AND OBJECTIVE: Complex fractionated atrial electrograms (CFAE) may contain information concerning the electrophysiological substrate of atrial fibrillation (AF); therefore they are of interest to guide catheter ablation treatment of AF. Electrogram signals are shaped by activation events, which are dynamical in nature. This makes it difficult to establish those signal properties that can provide insight into the ablation site location. Nonlinear measures may improve information. To test this hypothesis, we used nonlinear measures to analyze CFAE. METHODS: CFAE from several atrial sites, recorded for a duration of 16 s, were acquired from 10 patients with persistent and 9 patients with paroxysmal AF. These signals were appraised using non-overlapping windows of 1-, 2- and 4-s durations. The resulting data sets were analyzed with Recurrence Plots (RP) and Recurrence Quantification Analysis (RQA). The data was also quantified via entropy measures. RESULTS: RQA exhibited unique plots for persistent versus paroxysmal AF. Similar patterns were observed to be repeated throughout the RPs. Trends were consistent for signal segments of 1 and 2 s as well as 4 s in duration. This was suggestive that the underlying signal generation process is also repetitive, and that repetitiveness can be detected even in 1-s sequences. The results also showed that most entropy metrics exhibited higher measurement values (closer to equilibrium) for persistent AF data. It was also found that Determinism (DET), Trapping Time (TT), and Modified Multiscale Entropy (MMSE), extracted from signals that were acquired from locations at the posterior atrial free wall, are highly discriminative of persistent versus paroxysmal AF data. CONCLUSIONS: Short data sequences are sufficient to provide information to discern persistent versus paroxysmal AF data with a significant difference, and can be useful to detect repeating patterns of atrial activation.

Cerebrovascular pattern improved by ozone autohemotherapy: an entropy-based study on multiple sclerosis patients.

Ozone major autohemotherapy is effective in reducing the symptoms of multiple sclerosis (MS) patients, but its effects on brain are still not clear. In this work, we have monitored the changes in the cerebrovascular pattern of MS patients and normal subjects during major ozone autohemotherapy by using near-infrared spectroscopy (NIRS) as functional and vascular technique. NIRS signals are analyzed using a combination of time, time-frequency analysis and nonlinear analysis of intrinsic mode function signals obtained from empirical mode decomposition technique. Our results show that there is an improvement in the cerebrovascular pattern of all subjects indicated by increasing the entropy of the NIRS signals. Hence, we can conclude that the ozone therapy increases the brain metabolism and helps to recover from the lower activity levels which is predominant in MS patients.

Longitudinal changes in tear fluid lipidome brought about by eyelid-warming treatment in a cohort of meibomian gland dysfunction.

Meibomian gland dysfunction (MGD) is a leading cause of evaporative dry eye and ocular discomfort characterized by an unstable tear film principally attributed to afflicted delivery of lipids to the ocular surface. Herein, we elucidated longitudinal tear lipid alterations associated with disease alleviation and symptom improvement in a cohort of MGD patients undergoing eyelid-warming treatment for 12 weeks. Remarkably, eyelid-warming resulted in stark reductions in lysophospholipids (P < 0.001 for lyso-plasmalogen phosphatidylethanolamine, lysophosphatidylcholine, and lysophosphatidylinositol), as well as numerous PUFA-containing diacylglyceride species in tears, accompanied by significant increases in several PUFA-containing phospholipids. These changes in tear lipidomes suggest that eyelid-warming leads to diminished activity of tear phospholipases that preferentially target PUFA-containing phospholipids. In addition, treatment led to appreciable increases (P < 0.001) in O-acyl-ω-hydroxy-FAs (OAHFAs), which are lipid amphiphiles critical to the maintenance of tear film stability. Longitudinal changes in the tear lipids aforementioned also significantly (P < 0.05) correlated with reduced rate of ocular evaporation and improvement in ocular symptoms. The foregoing data thus indicate that excess ocular surface phospholipase activity detrimental to tear film stability could be alleviated by eyelid warming alone without application of steroids and identify tear OAHFAs as suitable markers to monitor treatment response in MGD.

Data mining framework for fatty liver disease classification in ultrasound: a hybrid feature extraction paradigm.

PURPOSE: Fatty liver disease (FLD) is an increasing prevalent disease that can be reversed if detected early. Ultrasound is the safest and ubiquitous method for identifying FLD. Since expert sonographers are required to accurately interpret the liver ultrasound images, lack of the same will result in interobserver variability. For more objective interpretation, high accuracy, and quick second opinions, computer aided diagnostic (CAD) techniques may be exploited. The purpose of this work is to develop one such CAD technique for accurate classification of normal livers and abnormal livers affected by FLD. METHODS: In this paper, the authors present a CAD technique (called Symtosis) that uses a novel combination of significant features based on the texture, wavelet transform, and higher order spectra of the liver ultrasound images in various supervised learning-based classifiers in order to determine parameters that classify normal and FLD-affected abnormal livers. RESULTS: On evaluating the proposed technique on a database of 58 abnormal and 42 normal liver ultrasound images, the authors were able to achieve a high classification accuracy of 93.3% using the decision tree classifier. CONCLUSIONS: This high accuracy added to the completely automated classification procedure makes the authors' proposed technique highly suitable for clinical deployment and usage.

Probabilistic information structure of human walking.

Recently, the area of healthcare has been tremendously benefited from the advent of high performance computing in improving quality of life. Different processing techniques have been developed to understand the hidden complexity of the time series and will help clinicians in diagnosis and treatment. Analysis of human walking helps to study the various pathological conditions affecting balance and the elderly. In an elderly subjects, falls and paralysis are major problems, in terms of both frequency and consequences. Correct postural balance is important to well being and its effects will be felt in every movement and activity. In this paper, Bayesian Network (BN) was applied to recorded muscle activities and joint motions during walking, to extract causal information structure of normal walking and different impaired walking. The aim of this study is to use different BNs to express normal walking and various impaired walking, and identify the most important causal pairs that characterize specific impaired walking, through comparing the BNs for different walking.

Towards the systematic development of medical networking technology.

Currently, there is a disparity in the availability of doctors between urban and rural areas of developing countries. Most experienced doctors and specialists, as well as advanced diagnostic technologies, are available in urban areas. People living in rural areas have less or sometimes even no access to affordable healthcare facilities. Increasing the number of doctors and charitable medical hospitals or deploying advanced medical technologies in these areas might not be economically feasible, especially in developing countries. We need to mobilize science and technology to master this complex, large scale problem in an objective, logical, and professional way. This can only be achieved with a collaborative effort where a team of experts works on both technical and non-technical aspects of this health care divide. In this paper we use a systems engineering framework to discuss hospital networks which might be solution for the problem. We argue that with the advancement in communication and networking technologies, economically middle class people and even some rural poor have access to internet and mobile communication systems. Thus, Hospital Digital Networking Technologies (HDNT), such as telemedicine, can be developed to utilize internet, mobile and satellite communication systems to connect primitive rural healthcare centers to well advanced modern urban setups and thereby provide better consultation and diagnostic care to the needy people. This paper describes requirements and limitations of the HDNTs. It also presents the features of telemedicine, the implementation issues and the application of wireless technologies in the field of medical networking.

EEG signal analysis: a survey.

The EEG (Electroencephalogram) signal indicates the electrical activity of the brain. They are highly random in nature and may contain useful information about the brain state. However, it is very difficult to get useful information from these signals directly in the time domain just by observing them. They are basically non-linear and nonstationary in nature. Hence, important features can be extracted for the diagnosis of different diseases using advanced signal processing techniques. In this paper the effect of different events on the EEG signal, and different signal processing methods used to extract the hidden information from the signal are discussed in detail. Linear, Frequency domain, time - frequency and non-linear techniques like correlation dimension (CD), largest Lyapunov exponent (LLE), Hurst exponent (H), different entropies, fractal dimension(FD), Higher Order Spectra (HOS), phase space plots and recurrence plots are discussed in detail using a typical normal EEG signal.

Nonlinear analysis of EEG signals at different mental states.

BACKGROUND: The EEG (Electroencephalogram) is a representative signal containing information about the condition of the brain. The shape of the wave may contain useful information about the state of the brain. However, the human observer can not directly monitor these subtle details. Besides, since bio-signals are highly subjective, the symptoms may appear at random in the time scale. Therefore, the EEG signal parameters, extracted and analyzed using computers, are highly useful in diagnostics. This work discusses the effect on the EEG signal due to music and reflexological stimulation. METHODS: In this work, nonlinear parameters like Correlation Dimension (CD), Largest Lyapunov Exponent (LLE), Hurst Exponent (H) and Approximate Entropy (ApEn) are evaluated from the EEG signals under different mental states. RESULTS: The results obtained show that EEG to become less complex relative to the normal state with a confidence level of more than 85% due to stimulation. CONCLUSIONS: It is found that the measures are significantly lower when the subjects are under sound or reflexologic stimulation as compared to the normal state. The dimension increases with the degree of the cognitive activity. This suggests that when the subjects are under sound or reflexologic stimuli, the number of parallel functional processes active in the brain is less and the brain goes to a more relaxed state