Explainable Model Using Shapley Additive Explanations Approach on Wound Infection after Wide Soft Tissue Sarcoma Resection: "Big Data" Analysis Based on Health Insurance Review and Assessment Service Hub.

Background and Objectives: Soft tissue sarcomas represent a heterogeneous group of malignant mesenchymal tissues. Despite their low prevalence, soft tissue sarcomas present clinical challenges for orthopedic surgeons owing to their aggressive nature, and perioperative wound infections. However, the low prevalence of soft tissue sarcomas has hindered the availability of large-scale studies. This study aimed to analyze wound infections after wide resection in patients with soft tissue sarcomas by employing big data analytics from the Hub of the Health Insurance Review and Assessment Service (HIRA). Materials and Methods: Patients who underwent wide excision of soft tissue sarcomas between 2010 and 2021 were included. Data were collected from the HIRA database of approximately 50 million individuals' information in the Republic of Korea. The data collected included demographic information, diagnoses, prescribed medications, and surgical procedures. Random forest has been used to analyze the major associated determinants. A total of 10,906 observations with complete data were divided into training and validation sets in an 80:20 ratio (8773 vs. 2193 cases). Random forest permutation importance was employed to identify the major predictors of infection and Shapley Additive Explanations (SHAP) values were derived to analyze the directions of associations with predictors. Results: A total of 10,969 patients who underwent wide excision of soft tissue sarcomas were included. Among the study population, 886 (8.08%) patients had post-operative infections requiring surgery. The overall transfusion rate for wide excision was 20.67% (2267 patients). Risk factors among the comorbidities of each patient with wound infection were analyzed and dependence plots of individual features were visualized. The transfusion dependence plot reveals a distinctive pattern, with SHAP values displaying a negative trend for individuals without blood transfusions and a positive trend for those who received blood transfusions, emphasizing the substantial impact of blood transfusions on the likelihood of wound infection. Conclusions: Using the machine learning random forest model and the SHAP values, the perioperative transfusion, male sex, old age, and low SES were important features of wound infection in soft-tissue sarcoma patients.

Real-time Classification of Fetal Status Based on Deep Learning and Cardiotocography Data.

This study uses convolutional neural networks (CNNs) and cardiotocography data for the real-time classification of fetal status in the mobile application of a pregnant woman and the computer server of a data expert at the same time (The sensor is connected with the smartphone, which is linked with the web server for the woman and the computer server for the expert). Data came from 5249 (or 4833) cardiotocography traces in Anam Hospital for the mobile application (or the computer server). 150 data cases of 5-minute duration were extracted from each trace with 141,001 final cases for the mobile application and for the computer server alike. The dependent variable was fetal status with two categories (Normal, Abnormal) for the mobile application and three categories (Normal, Middle, Abnormal) for the computer server. The fetal heart rate served as a predictor for the mobile application and the computer server, while uterus contraction for the computer server only. The 1-dimension (or 2-dimension) Resnet CNN was trained for the mobile application (or the computer server) during 800 epochs. The sensitivity, specificity and their harmonic mean of the 1-dimension CNN for the mobile application were 94.9%, 91.2% and 93.0%, respectively. The corresponding statistics of the 2-dimension CNN for the computer server were 98.0%, 99.5% and 98.7%. The average inference time per 1000 images was 6.51 micro-seconds. Deep learning provides an efficient model for the real-time classification of fetal status in the mobile application and the computer server at the same time.

Radiomic machine learning for predicting prognostic biomarkers and molecular subtypes of breast cancer using tumor heterogeneity and angiogenesis properties on MRI.

OBJECTIVES: To investigate machine learning approaches for radiomics-based prediction of prognostic biomarkers and molecular subtypes of breast cancer using quantification of tumor heterogeneity and angiogenesis properties on magnetic resonance imaging (MRI). METHODS: This prospective study examined 291 invasive cancers in 288 patients who underwent breast MRI at 3 T before treatment between May 2017 and July 2019. Texture and perfusion analyses were performed and a total of 160 parameters for each cancer were extracted. Relationships between MRI parameters and prognostic biomarkers were analyzed using five machine learning algorithms. Each model was built using only texture features, only perfusion features, or both. Model performance was compared using the area under the receiver-operating characteristic curve (AUC) and the DeLong method, and the importance of MRI parameters in prediction was derived. RESULTS: Texture parameters were associated with the status of hormone receptors, human epidermal growth factor receptor 2, and Ki67, tumor size, grade, and molecular subtypes (p < 0.002). Perfusion parameters were associated with the status of hormone receptors and Ki67, grade, and molecular subtypes (p < 0.003). The random forest model integrating texture and perfusion parameters showed the highest performance (AUC = 0.75). The performance of the random forest model was the best with a special scale filter of 0 (AUC = 0.80). The important parameters for prediction were texture irregularity (entropy) and relative extracellular extravascular space (Ve). CONCLUSIONS: Radiomic machine learning that integrates tumor heterogeneity and angiogenesis properties on MRI has the potential to noninvasively predict prognostic factors of breast cancer. KEY POINTS: • Machine learning, integrating tumor heterogeneity and angiogenesis properties on MRI, can be applied to predict prognostic biomarkers and molecular subtypes in breast cancer. • The random forest model showed the best predictive performance among the five machine learning models (logistic regression, decision tree, naïve Bayes, random forest, and artificial neural network). • The most important MRI parameters for predicting prognostic factors in breast cancer were texture irregularity (entropy) among texture parameters and relative extracellular extravascular space (Ve) among perfusion parameters.

Association of preterm birth with medications: machine learning analysis using national health insurance data.

PURPOSE: To use machine learning and population data for testing the associations of preterm birth with socioeconomic status, gastroesophageal reflux disease (GERD) and medication history including proton pump inhibitors, sleeping pills and antidepressants. METHODS: Population-based retrospective cohort data came from Korea National Health Insurance Service claims data for all women who aged 25-40 years and gave births for the first time as singleton pregnancy during 2015-2017 (405,586 women). The dependent variable was preterm birth during 2015-2017 and 65 independent variables were included (demographic/socioeconomic determinants, disease information, medication history, obstetric information). Random forest variable importance (outcome measure) was used for identifying major determinants of preterm birth and testing its associations with socioeconomic status, GERD and medication history including proton pump inhibitors, sleeping pills and antidepressants. RESULTS: Based on random forest variable importance, major determinants of preterm birth during 2015-2017 were socioeconomic status (645.34), age (556.86), proton pump inhibitors (107.61), GERD for the years 2014, 2012 and 2013 (106.78, 105.87 and 104.96), sleeping pills (97.23), GERD for the years 2010, 2011 and 2009 (95.56, 94.84 and 93.81), and antidepressants (90.13). CONCLUSION: Preterm birth has strong associations with low socioeconomic status, GERD and medication history such as proton pump inhibitors, sleeping pills and antidepressants. For preventing preterm birth, appropriate medication would be needed alongside preventive measures for GERD and the promotion of socioeconomic status for pregnant women.

Compressed Deep Learning to Classify Arrhythmia in an Embedded Wearable Device.

The importance of an embedded wearable device with automatic detection and alarming cannot be overstated, given that 15-30% of patients with atrial fibrillation are reported to be asymptomatic. These asymptomatic patients do not seek medical care, hence traditional diagnostic tools including Holter are not effective for the further prevention of associated stroke or heart failure. This is likely to be more so in the era of COVID-19, in which patients become more reluctant on hospitalization and checkups. However, little literature is available on this important topic. For this reason, this study developed efficient deep learning with model compression, which is designed to use ECG data and classify arrhythmia in an embedded wearable device. ECG-signal data came from Korea University Anam Hospital in Seoul, Korea, with 28,308 unique patients (15,412 normal and 12,896 arrhythmia). Resnets and Mobilenets with model compression (TensorFlow Lite) were applied and compared for the diagnosis of arrhythmia in an embedded wearable device. The weight size of the compressed model registered a remarkable decrease from 743 MB to 76 KB (1/10000), whereas its performance was almost the same as its original counterpart. Resnet and Mobilenet were similar in terms of accuracy, i.e., Resnet-50 Hz (97.3) vs. Mo-bilenet-50 Hz (97.2), Resnet-100 Hz (98.2) vs. Mobilenet-100 Hz (97.9). Here, 50 Hz/100 Hz denotes the down-sampling rate. However, Resnets took more flash memory and longer inference time than did Mobilenets. In conclusion, Mobilenet would be a more efficient model than Resnet to classify arrhythmia in an embedded wearable device.

Artificial intelligence approaches to the determinants of women's vaginal dryness using general hospital data.

The aim of this study is to analyse the determinants of women's vaginal dryness using machine learning. Data came from Korea University Anam Hospital in Seoul, Republic of Korea, with 3298 women, aged 40-80 years, who attended their general health check from January 2010 to December 2012. Five machine learning methods were applied and compared for the prediction of vaginal dryness, measured by a Menopause Rating Scale. Random forest variable importance, a performance gap between a complete model and a model excluding a certain variable, was adopted for identifying major determinants of vaginal dryness. In terms of the mean squared error, the random forest (1.0597) was much better than linear regression (17.9043) and artificial neural networks with one, two and three hidden layers (1.7452, 1.7148 and 1.7736, respectively). Based on random forest variable importance, the top-10 determinants of vaginal dryness were menopause age, age, menopause, height, thyroid stimulating hormone, neutrophils, years since menopause, lymphocytes, alkaline phosphatase and blood urea nitrogen. In addition, its top-20 determinants were peak expiratory flow rate, low-density lipoprotein cholesterol, white blood cells, monocytes, cancer antigen 19-9, creatinine, eosinophils, total cholesterol, triglyceride and amylase. Machine learning presents a great decision support system for the prediction of vaginal dryness. For preventing vaginal dryness, preventive measures would be needed regarding early menopause, the thyroid function and systematic inflammation.Impact StatementWhat is already known on this subject? Only a few studies have investigated the risk factors of vaginal dryness in middle-aged women. More research is to be done for finding its various risk factors, identifying its major risk groups and drawing its effective clinical implications.What do the results of this study add? This study is the first machine-learning study to predict women's vaginal dryness and analyse their determinants. The random forest could discuss which factors are more important for the prediction of vaginal dryness. Based on random forest variable importance, menopause age was the most important determinant of vaginal dryness and their association was discovered to be negative in this study. Vaginal dryness was closely associated with the height, rather than the body weight or body mass index. The importance rankings of blood conditions related to systematic inflammation were within the top-20 in this study: neutrophils, lymphocytes, white blood cells, monocytes and eosinophils.What are the implications of these findings for clinical practice and/or further research? Machine learning presents a great decision support system for the prediction of vaginal dryness. For preventing vaginal dryness, preventive measures would be needed regarding early menopause and systematic inflammation.

Prediction of newborn's body mass index using nationwide multicenter ultrasound data: a machine-learning study.

BACKGROUND: This study introduced machine learning approaches to predict newborn's body mass index (BMI) based on ultrasound measures and maternal/delivery information. METHODS: Data came from 3159 obstetric patients and their newborns enrolled in a multi-center retrospective study. Variable importance, the effect of a variable on model performance, was used for identifying major predictors of newborn's BMI among ultrasound measures and maternal/delivery information. The ultrasound measures included biparietal diameter (BPD), abdominal circumference (AC) and estimated fetal weight (EFW) taken three times during the week 21 - week 35 of gestational age and once in the week 36 or later. RESULTS: Based on variable importance from the random forest, major predictors of newborn's BMI were the first AC and EFW in the week 36 or later, gestational age at delivery, the first AC during the week 21 - the week 35, maternal BMI at delivery, maternal weight at delivery and the first BPD in the week 36 or later. For predicting newborn's BMI, linear regression (2.0744) and the random forest (2.1610) were better than artificial neural networks with one, two and three hidden layers (150.7100, 154.7198 and 152.5843, respectively) in the mean squared error. CONCLUSIONS: This is the first machine-learning study with 64 clinical and sonographic markers for the prediction of newborns' BMI. The week 36 or later is the most effective period for taking the ultrasound measures and AC and EFW are the best predictors of newborn's BMI alongside gestational age at delivery and maternal BMI at delivery.

Automated detection of colorectal tumors based on artificial intelligence.

BACKGROUND: This study developed a diagnostic tool to automatically detect normal, unclear and tumor images from colonoscopy videos using artificial intelligence. METHODS: For the creation of training and validation sets, 47,555 images in the jpg format were extracted from colonoscopy videos for 24 patients in Korea University Anam Hospital. A gastroenterologist with the clinical experience of 15 years divided the 47,555 images into three classes of Normal (25,895), Unclear (2038) and Tumor (19,622). A single shot detector, a deep learning framework designed for object detection, was trained using the 47,255 images and validated with two sets of 300 images-each validation set included 150 images (50 normal, 50 unclear and 50 tumor cases). Half of the 47,255 images were used for building the model and the other half were used for testing the model. The learning rate of the model was 0.0001 during 250 epochs (training cycles). RESULTS: The average accuracy, precision, recall, and F1 score over the category were 0.9067, 0.9744, 0.9067 and 0.9393, respectively. These performance measures had no change with respect to the intersection-over-union threshold (0.45, 0.50, and 0.55). This finding suggests the stability of the model. CONCLUSION: Automated detection of normal, unclear and tumor images from colonoscopy videos is possible by using a deep learning framework. This is expected to provide an invaluable decision supporting system for clinical experts.

Association of Gastroesophageal Reflux Disease with Preterm Birth: Machine Learning Analysis.

BACKGROUND: This study used machine learning and population data for testing the associations of preterm birth with gastroesophageal reflux disease (GERD) and periodontitis. METHODS: Retrospective cohort data came from Korea National Health Insurance Service claims data for all women who aged 25-40 years and gave births for the first time as singleton pregnancy during 2015-2017 (405,586 women). The dependent variable was preterm birth during 2015-2017 and the independent variables were GERD (coded as no vs. yes) for each of the years 2002-2014, periodontitis (coded as no vs. yes) for each of the years 2002-2014, age (year) in 2014, socioeconomic status in 2014 measured by an insurance fee, and region (city) (coded as no vs. yes) in 2014. Random forest variable importance was adopted for finding main predictors of preterm birth and testing its associations with GERD and periodontitis. RESULTS: Based on random forest variable importance, main predictors of preterm birth during 2015-2017 were socioeconomic status in 2014, age in 2014, GERD for the years 2012, 2014, 2010, 2013, 2007 and 2009, region (city) in 2014 and GERD for the year 2006. The importance rankings of periodontitis were relatively low. CONCLUSION: Preterm birth has a stronger association with GERD than with periodontitis. For the prevention of preterm birth, preventive measures for GERD would be essential together with the improvement of socioeconomic status for pregnant women. Especially, it would be vital to promote active counseling for general GERD symptoms (neglected by pregnant women).

Machine Learning Approaches to Identify Factors Associated with Women's Vasomotor Symptoms Using General Hospital Data.

BACKGROUND: To analyze the factors associated with women's vasomotor symptoms (VMS) using machine learning. METHODS: Data on 3,298 women, aged 40-80 years, who attended their general health check-up from January 2010 to December 2012 were obtained from Korea University Anam Hospital in Seoul, Korea. Five machine learning methods were applied and compared for the prediction of VMS, measured by the Menopause Rating Scale. Variable importance, the effect of a variable on model performance, was used for identifying the major factors associated with VMS. RESULTS: In terms of the mean squared error, the random forest (0.9326) was much better than linear regression (12.4856) and artificial neural networks with one, two, and three hidden layers (1.5576, 1.5184, and 1.5833, respectively). Based on the variable importance from the random forest, the most important factors associated with VMS were age, menopause age, thyroid-stimulating hormone, and monocyte, triglyceride, gamma glutamyl transferase, blood urea nitrogen, cancer antigen 19-9, C-reactive protein, and low-density lipoprotein cholesterol levels. Indeed, the following variables were ranked within the top 20 in terms of variable importance: cancer antigen 125, total cholesterol, insulin, free thyroxine, forced vital capacity, alanine aminotransferase, forced expired volume in 1 second, height, homeostatic model assessment for insulin resistance, and carcinoembryonic antigen. CONCLUSION: Machine learning provides an invaluable decision support system for the prediction of VMS. For managing VMS, comprehensive consideration is needed regarding thyroid function, lipid profile, liver function, inflammation markers, insulin resistance, monocyte count, cancer antigens, and lung function.

Determinants of Spontaneous Preterm Labor and Birth Including Gastroesophageal Reflux Disease and Periodontitis.

BACKGROUND: Periodontitis is reported to be associated with preterm birth (spontaneous preterm labor and birth). Gastroesophageal reflux disease (GERD) is common during pregnancy and is expected to be related to periodontitis. However, little research has been done on the association among preterm birth, GERD and periodontitis. This study uses popular machine learning methods for analyzing preterm birth, GERD and periodontitis. METHODS: Data came from Anam Hospital in Seoul, Korea, with 731 obstetric patients during January 5, 1995 - August 28, 2018. Six machine learning methods were applied and compared for the prediction of preterm birth. Variable importance, the effect of a variable on model performance, was used for identifying major determinants of preterm birth. RESULTS: In terms of accuracy, the random forest (0.8681) was similar with logistic regression (0.8736). Based on variable importance from the random forest, major determinants of preterm birth are delivery and pregestational body mass indexes (BMI) (0.1426 and 0.1215), age (0.1211), parity (0.0868), predelivery systolic and diastolic blood pressure (0.0809 and 0.0763), twin (0.0476), education (0.0332) as well as infant sex (0.0331), prior preterm birth (0.0290), progesterone medication history (0.0279), upper gastrointestinal tract symptom (0.0274), GERD (0.0242), Helicobacter pylori (0.0151), region (0.0139), calcium-channel-blocker medication history (0.0135) and gestational diabetes mellitus (0.0130). Periodontitis ranked 22nd (0.0084). CONCLUSION: GERD is more important than periodontitis for predicting and preventing preterm birth. For preventing preterm birth, preventive measures for hypertension, GERD and diabetes mellitus would be needed alongside the promotion of effective BMI management and appropriate progesterone and calcium-channel-blocker medications.

Artificial Neural Network Analysis of Spontaneous Preterm Labor and Birth and Its Major Determinants.

BACKGROUND: Little research based on the artificial neural network (ANN) is done on preterm birth (spontaneous preterm labor and birth) and its major determinants. This study uses an ANN for analyzing preterm birth and its major determinants. METHODS: Data came from Anam Hospital in Seoul, Korea, with 596 obstetric patients during March 27, 2014 - August 21, 2018. Six machine learning methods were applied and compared for the prediction of preterm birth. Variable importance, the effect of a variable on model performance, was used for identifying major determinants of preterm birth. Analysis was done in December, 2018. RESULTS: The accuracy of the ANN (0.9115) was similar with those of logistic regression and the random forest (0.9180 and 0.8918, respectively). Based on variable importance from the ANN, major determinants of preterm birth are body mass index (0.0164), hypertension (0.0131) and diabetes mellitus (0.0099) as well as prior cone biopsy (0.0099), prior placenta previa (0.0099), parity (0.0033), cervical length (0.0001), age (0.0001), prior preterm birth (0.0001) and myomas & adenomyosis (0.0001). CONCLUSION: For preventing preterm birth, preventive measures for hypertension and diabetes mellitus are required alongside the promotion of cervical-length screening with different guidelines across the scope/type of prior conization.

Atrial fibrillation classification based on convolutional neural networks.

BACKGROUND: The global age-adjusted mortality rate related to atrial fibrillation (AF) registered a rapid growth in the last four decades, i.e., from 0.8 to 1.6 and 0.9 to 1.7 per 100,000 for men and women during 1990-2010, respectively. In this context, this study uses convolutional neural networks for classifying (diagnosing) AF, employing electrocardiogram data in a general hospital. METHODS: Data came from Anam Hospital in Seoul, Korea, with 20,000 unique patients (10,000 normal sinus rhythm and 10,000 AF). 30 convolutional neural networks were applied and compared for the diagnosis of the normal sinus rhythm vs. AF condition: 6 Alex networks with 5 convolutional layers, 3 fully connected layers and the number of kernels changing from 3 to 256; and 24 residual networks with the number of residuals blocks (or kernels) varying from 8 to 2 (or 64 to 2). RESULTS: In terms of the accuracy, the best Alex network was one with 24 initial kernels (i.e., kernels in the first layer), 5,268,818 parameters and the training time of 89 s (0.997), while the best residual network was one with 6 residual blocks, 32 initial kernels, 248,418 parameters and the training time of 253 s (0.999). In general, the performance of the residual network improved as the number of its residual blocks (its depth) increased. CONCLUSION: For AF diagnosis, the residual network might be a good model with higher accuracy and fewer parameters than its Alex-network counterparts.

A change in social activity and depression among Koreans aged 45 years and more: analysis of the Korean Longitudinal Study of Aging (2006-2010).

BACKGROUND: The aim of this study is to examine a relationship between a change in social activity and depression among Koreans aged 45 years or more. METHODS: Data came from the Korean Longitudinal Study of Aging (KLoSA) (2006-2010), with 5,327 participants aged 45 years or more. The generalized estimating equation (GEE) with the logit link was used to investigate an association between a change in social activity during 2006-2008 (or 2008-2010) and depression among respondents in year 2008 (or Y2010). Depression was measured by Center for Epidemiological Studies Depression scale (CES-D10) and a change in social activity was classified with four categories, i.e. "consistent participation", "consistent non-participation", "participation to non-participation", and "non-participation to participation". Social activity was divided into various elements and the same analysis was conducted for each of these elements. RESULTS: Those with consistent non-participation and from participation to non-participation were more likely to be depressed than those with consistent participation and from non-participation to participation in social activities (OR 1.44 [95% CI 1.22-1.71], OR 1.35 [95% CI 1.15-1.58] vs. OR 1.00 [Reference], OR 1.27 [95% CI 1.09-1.48]). In addition, the strength of the negative association between consistent or new participation in social activity and depression was different across different elements of social activity. The negative association was particularly strong for leisure, culture or sports clubs, and for family or school reunion. CONCLUSION: For improving the mental health of the population aged 45 years or more, the promotion of their continued or new participations in leisure/culture clubs and family/school reunion might be needed in South Korea.

Artificial intelligence in colonoscopy: from detection to diagnosis.

This study reviews the recent progress of artificial intelligence for colonoscopy from detection to diagnosis. The source of data was 27 original studies in PubMed. The search terms were "colonoscopy" (title) and "deep learning" (abstract). The eligibility criteria were: (1) the dependent variable of gastrointestinal disease; (2) the interventions of deep learning for classification, detection and/or segmentation for colonoscopy; (3) the outcomes of accuracy, sensitivity, specificity, area under the curve (AUC), precision, F1, intersection of union (IOU), Dice and/or inference frames per second (FPS); (3) the publication year of 2021 or later; (4) the publication language of English. Based on the results of this study, different deep learning methods would be appropriate for different tasks for colonoscopy, e.g., Efficientnet with neural architecture search (AUC 99.8%) in the case of classification, You Only Look Once with the instance tracking head (F1 96.3%) in the case of detection, and Unet with dense-dilation-residual blocks (Dice 97.3%) in the case of segmentation. Their performance measures reported varied within 74.0-95.0% for accuracy, 60.0-93.0% for sensitivity, 60.0-100.0% for specificity, 71.0-99.8% for the AUC, 70.1-93.3% for precision, 81.0-96.3% for F1, 57.2-89.5% for the IOU, 75.1-97.3% for Dice and 66-182 for FPS. In conclusion, artificial intelligence provides an effective, non-invasive decision support system for colonoscopy from detection to diagnosis.

Clinical and dental predictors of preterm birth using machine learning methods: the MOHEPI study.

Preterm birth (PTB) is one of the most common and serious complications of pregnancy, leading to mortality and severe morbidities that can impact lifelong health. PTB could be associated with various maternal medical condition and dental status including periodontitis. The purpose of this study was to identify major predictors of PTB among clinical and dental variables using machine learning methods. Prospective cohort data were obtained from 60 women who delivered singleton births via cesarean section (30 PTB, 30 full-term birth [FTB]). Dependent variables were PTB and spontaneous PTB (SPTB). 15 independent variables (10 clinical and 5 dental factors) were selected for inclusion in the machine learning analysis. Random forest (RF) variable importance was used to identify the major predictors of PTB and SPTB. Shapley additive explanation (SHAP) values were calculated to analyze the directions of the associations between the predictors and PTB/SPTB. Major predictors of PTB identified by RF variable importance included pre-pregnancy body mass index (BMI), modified gingival index (MGI), preeclampsia, decayed missing filled teeth (DMFT) index, and maternal age as in top five rankings. SHAP values revealed positive correlations between PTB/SPTB and its major predictors such as premature rupture of the membranes, pre-pregnancy BMI, maternal age, and MGI. The positive correlations between these predictors and PTB emphasize the need for integrated medical and dental care during pregnancy. Future research should focus on validating these predictors in larger populations and exploring interventions to mitigate these risk factors.

Prediction of gait recovery using machine learning algorithms in patients with spinal cord injury.

With advances in artificial intelligence, machine learning (ML) has been widely applied to predict functional outcomes in clinical medicine. However, there has been no attempt to predict walking ability after spinal cord injury (SCI) based on ML. In this situation, the main purpose of this study was to predict gait recovery after SCI at discharge from an acute rehabilitation facility using various ML algorithms. In addition, we explored important variables that were related to the prognosis. Finally, we attempted to suggest an ML-based decision support system (DSS) for predicting gait recovery after SCI. Data were collected retrospectively from patients with SCI admitted to an acute rehabilitation facility between June 2008 to December 2021. Linear regression analysis and ML algorithms (random forest [RF], decision tree [DT], and support vector machine) were used to predict the functional ambulation category at the time of discharge (FAC_DC) in patients with traumatic or non-traumatic SCI (n = 353). The independent variables were age, sex, duration of acute care and rehabilitation, comorbidities, neurological information entered into the International Standards for Neurological Classification of SCI worksheet, and somatosensory-evoked potentials at the time of admission to the acute rehabilitation facility. In addition, the importance of variables and DT-based DSS for FAC_DC was analyzed. As a result, RF and DT accurately predicted the FAC_DC measured by the root mean squared error. The root mean squared error of RF and the DT were 1.09 and 1.24 for all participants, 1.20 and 1.06 for those with trauma, and 1.12 and 1.03 for those with non-trauma, respectively. In the analysis of important variables, the initial FAC was found to be the most influential factor in all groups. In addition, we could provide a simple DSS based on strong predictors such as the initial FAC, American Spinal Injury Association Impairment Scale grades, and neurological level of injury. In conclusion, we provide that ML can accurately predict gait recovery after SCI for the first time. By focusing on important variables and DSS, we can guide early prognosis and establish personalized rehabilitation strategies in acute rehabilitation hospitals.

Explainable artificial intelligence on safe balance and its major determinants in stroke patients.

This study develops explainable artificial intelligence for predicting safe balance using hospital data, including clinical, neurophysiological, and diffusion tensor imaging properties. Retrospective data from 92 first-time stroke patients from January 2016 to June 2023 was analysed. The dependent variables were independent mobility scores, i.e., Berg Balance Scales with 0 (45 or below) vs. 1 (above 45) measured after three and six months, respectively. Twenty-nine predictors were included. Random forest variable importance was employed for identifying significant predictors of the Berg Balance Scale and testing its associations with the predictors, including Berg Balance Scale after one month and corticospinal tract diffusion tensor imaging properties. Shapley Additive Explanation values were calculated to analyse the directions of these associations. The random forest registered a higher or similar area under the curve compared to logistic regression, i.e., 91% vs. 87% (Berg Balance Scale after three months), 92% vs. 92% (Berg Balance Scale after six months). Based on random forest variable importance values and rankings: (1) Berg Balance Scale after three months has strong associations with Berg Balance Scale after one month, Fugl-Meyer assessment scale, ipsilesional corticospinal tract fractional anisotropy, fractional anisotropy laterality index and age; (2) Berg Balance Scale after six months has strong relationships with Fugl-Meyer assessment scale, Berg Balance Scale after one month, ankle plantar flexion muscle strength, knee extension muscle strength and hip flexion muscle strength. These associations were positive in the SHAP summary plots. Including Berg Balance Scale after one month, Fugl-Meyer assessment scale or ipsilesional corticospinal tract fractional anisotropy in the random forest will increase the probability of Berg Balance Scale after three months being above 45 by 0.11, 0.08, or 0.08. In conclusion, safe balance after stroke strongly correlates with its initial motor function, Fugl-Meyer assessment scale, and ipsilesional corticospinal tract fractional anisotropy. Diffusion tensor imaging information aids in developing explainable artificial intelligence for predicting safe balance after stroke.

Explainable artificial intelligence for predicting red blood cell transfusion in geriatric patients undergoing hip arthroplasty: Machine learning analysis using national health insurance data.

This study uses machine learning and population data to analyze major determinants of blood transfusion among patients with hip arthroplasty. Retrospective cohort data came from Korea National Health Insurance Service claims data for 19,110 patients aged 65 years or more with hip arthroplasty in 2019. The dependent variable was blood transfusion (yes vs no) in 2019 and its 31 predictors were included. Random forest variable importance and Shapley Additive Explanations were used for identifying major predictors and the directions of their associations with blood transfusion. The random forest registered the area under the curve of 73.6%. Based on random forest variable importance, the top-10 predictors were anemia (0.25), tranexamic acid (0.17), age (0.16), socioeconomic status (0.05), spinal anesthesia (0.05), general anesthesia (0.04), sex (female) (0.04), dementia (0.03), iron (0.02), and congestive heart failure (0.02). These predictors were followed by their top-20 counterparts including cardiovascular disease, statin, chronic obstructive pulmonary disease, diabetes mellitus, chronic kidney disease, peripheral vascular disease, liver disease, solid tumor, myocardial infarction and hypertension. In terms of max Shapley Additive Explanations values, these associations were positive, e.g., anemia (0.09), tranexamic acid (0.07), age (0.09), socioeconomic status (0.05), spinal anesthesia (0.05), general anesthesia (0.04), sex (female) (0.02), dementia (0.03), iron (0.04), and congestive heart failure (0.03). For example, the inclusion of anemia, age, tranexamic acid or spinal anesthesia into the random forest will increase the probability of blood transfusion among patients with hip arthroplasty by 9%, 7%, 9% or 5%. Machine learning is an effective prediction model for blood transfusion among patients with hip arthroplasty. The high-risk group with anemia, age and comorbid conditions need to be treated with tranexamic acid, iron and/or other appropriate interventions.

Deep Learning-Based Prediction Model for Gait Recovery after a Spinal Cord Injury.

Predicting gait recovery after a spinal cord injury (SCI) during an acute rehabilitation phase is important for planning rehabilitation strategies. However, few studies have been conducted on this topic to date. In this study, we developed a deep learning-based prediction model for gait recovery after SCI upon discharge from an acute rehabilitation facility. Data were collected from 405 patients with acute SCI admitted to the acute rehabilitation facility of Korea University Anam Hospital between June 2008 and December 2022. The dependent variable was Functional Ambulation Category at the time of discharge (FAC-DC). Seventy-one independent variables were selected from the existing literature: basic information, International Standards for Neurological Classification of SCI scores, neurogenic bladders, initial FAC, and somatosensory-evoked potentials of the lower extremity. Recurrent neural network (RNN), linear regression (LR), Ridge, and Lasso methods were compared for FAC-DC prediction in terms of the root-mean-squared error (RMSE). RNN variable importance, which is the RMSE gap between a complete RNN model and an RNN model excluding a certain variable, was used to evaluate the contribution of this variable. Based on the results of this study, the performance of the RNN was far better than that of LR, Ridge, and Lasso. The respective RMSEs were 0.3738, 2.2831, 1.3161, and 1.0246 for all the participants; 0.3727, 1.7176, 1.3914, and 1.3524 for those with trauma; and 0.3728, 1.7516, 1.1012, and 0.8889 for those without trauma. In terms of RNN variable importance, lower-extremity motor strength (right and left ankle dorsiflexors, right knee extensors, and left long toe extensors) and the neurological level of injury were ranked among the top five across the boards. Therefore, initial FAC was the seventh, third, and ninth most important predictor for all participants, those with trauma, and those without trauma, respectively. In conclusion, this study developed a deep learning-based prediction model with excellent performance for gait recovery after SCI at the time of discharge from an acute rehabilitation facility. This study also demonstrated the strength of deep learning as an explainable artificial intelligence method for identifying the most important predictors.