Correlations of Complete Blood Count with Alanine and Aspartate Transaminase in Chinese Subjects and Prediction Based on Back-Propagation Artificial Neural Network (BP-ANN)

Multidimensional Machine Learning for Assessing Parameters Associated With COVID-19 in Vietnam: Validation Study

BACKGROUND

Machine learning (ML) is a type of artificial intelligence strategy. Its algorithms are used on big data sets to see patterns, learn from their results, and perform tasks autonomously without being instructed on how to address problems. New diseases like COVID-19 provide important data for ML. Therefore, all relevant parameters should be explicitly quantified and modeled.

OBJECTIVE

The purpose of this study was to determine (1) the overall preclinical characteristics, (2) the cumulative cutoff values and risk ratios (RRs), and (3) the factors associated with COVID-19 severity in unidimensional and multidimensional analyses involving 2173 SARS-CoV-2 patients.

METHODS

The study population consisted of 2173 patients (1587 mild status [mild group] and asymptomatic patients, 377 moderate status patients [moderate group], and 209 severe status patients [severe group]). The status of the patients was recorded from September 2021 to March 2022. Two correlation tests, relative risk, and RR were used to eliminate unbalanced parameters and select the most remarkable parameters. The independent methods of hierarchical cluster analysis and k-means were used to classify parameters according to their r values. Finally, network analysis provided a 3-dimensional view of the results.

RESULTS

COVID-19 severity was significantly correlated with age (mild-moderate group: RR 4.19, 95% CI 3.58-4.95; P<.001), scoring index of chest x-ray (mild-moderate group: RR 3.29, 95% CI 2.76-3.92; P<.001; moderate-severe group: RR 3.03, 95% CI 2.4023-3.8314; P<.001), percentage of neutrophils (mild-moderate group: RR 3.18, 95% CI 2.73-3.70; P<.001; moderate-severe group: RR 3.32, 95% CI 2.6480-4.1529; P<.001), quantity of neutrophils (moderate-severe group: RR 3.15, 95% CI 2.6153-3.8025; P<.001), albumin (moderate-severe group: RR 0.46, 95% CI 0.3650-0.5752; P<.001), C-reactive protein (mild-moderate group: RR 3.4, 95% CI 2.91-3.97; P<.001), and ratio of lymphocytes (moderate-severe group: RR 0.34, 95% CI 0.2743-0.4210; P<.001). Significant inversion of correlations among the severity groups is important. Alanine transaminase and leucocytes showed a significant negative correlation (r=-1; P<.001) in the mild group and a significant positive correlation in the moderate group (r=1; P<.001). Transferrin and anion Cl showed a significant positive correlation (r=1; P<.001) in the mild group and a significant negative correlation in the moderate group (r=-0.59; P<.001). The clustering and network analysis showed that in the mild-moderate group, the closest neighbors of COVID-19 severity were ferritin and age. C-reactive protein, scoring index of chest x-ray, albumin, and lactate dehydrogenase were the next closest neighbors of these 3 factors. In the moderate-severe group, the closest neighbors of COVID-19 severity were ferritin, fibrinogen, albumin, quantity of lymphocytes, scoring index of chest x-ray, white blood cell count, lactate dehydrogenase, and quantity of neutrophils.

CONCLUSIONS

This multidimensional study in Vietnam showed possible correlations between several elements and COVID-19 severity to provide clinical reference markers for surveillance and diagnostic management.

Comparison of Diagnosis Accuracy between a Backpropagation Artificial Neural Network Model and Linear Regression in Digestive Disease Patients: an Empirical Research

Introduction

A Noninvasive diagnosis model for digestive diseases is the vital issue for the current clinical research. Our systematic review is aimed at demonstrating diagnosis accuracy between the BP-ANN algorithm and linear regression in digestive disease patients, including their activation function and data structure.

Methods

We reported the systematic review according to the PRISMA guidelines. We searched related articles from seven electronic scholarly databases for comparison of the diagnosis accuracy focusing on BP-ANN and linear regression. The characteristics, patient number, input/output marker, diagnosis accuracy, and results/conclusions related to comparison were extracted independently based on inclusion criteria.

Results

Nine articles met all the criteria and were enrolled in our review. Of those enrolled articles, the publishing year ranged from 1991 to 2017. The sample size ranged from 42 to 3222 digestive disease patients, and all of the patients showed comparable biomarkers between the BP-ANN algorithm and linear regression. According to our study, 8 literature demonstrated that the BP-ANN model is superior to linear regression in predicting the disease outcome based on AUROC results. One literature reported linear regression to be superior to BP-ANN for the early diagnosis of colorectal cancer.

Conclusion

The BP-ANN algorithm and linear regression both had high capacity in fitting the diagnostic model and BP-ANN displayed more prediction accuracy for the noninvasive diagnosis model of digestive diseases. We compared the activation functions and data structure between BP-ANN and linear regression for fitting the diagnosis model, and the data suggested that BP-ANN was a comprehensive recommendation algorithm.

Neural Network-Based Study about Correlation Model between TCM Constitution and Physical Examination Indexes Based on 950 Physical Examinees

Purpose

To establish the correlation model between Traditional Chinese Medicine (TCM) constitution and physical examination indexes by backpropagation neural network (BPNN) technology. A new method for the identification of TCM constitution in clinics is proposed, which is trying to solve the problem like shortage of TCM doctor, complicated process, low efficiency, and unfavorable application in the current TCM constitution identification methods.

Methods

The corresponding effective samples were formed by sorting out and classifying the original data which were collected from physical examination indexes and TCM constitution types of 950 physical examinees, who were examined at the affiliated hospital of Chengdu University of TCM. The BPNN algorithm was implemented using the C# programming language and Google's AI library. Then, the training group and the test (validation) group of the effective samples were, respectively, input into the algorithm, to complete the construction and validation of the target model.

Results

For all the correlation models built in this paper, the accuracy of the training group and the test group of entire physical examination indexes-constitutional-type network model, respectively, was 88% and 53%, and the error was 0.001. For the other network models, the accuracy of the learning group and the test group and error, respectively, was as follows: liver function (31%, 42%, and 11.7), renal function (41%, 38%, and 6.7), blood routine (56%, 42%, and 2.4), and urine routine (60%, 40%, and 2.6).

Conclusions

The more the physical examination indexes are used in training, the more accurate the network model is established to predict TCM constitution. The sample data used in this paper showed that there was a relatively strong correlation between TCM constitution and physical examination indexes. Construction of the correlation model between physical examination indexes and TCM constitution is a kind of study for the integration of Chinese and Western medicine, which provides a new approach for the identification of TCM constitution, and it may be expected to avoid the existing problem of TCM constitution identification at present.