Osteoporosis risk prediction for bone mineral density assessment of postmenopausal women using machine learning.

PURPOSE: A number of clinical decision tools for osteoporosis risk assessment have been developed to select postmenopausal women for the measurement of bone mineral density. We developed and validated machine learning models with the aim of more accurately identifying the risk of osteoporosis in postmenopausal women compared to the ability of conventional clinical decision tools. MATERIALS AND METHODS: We collected medical records from Korean postmenopausal women based on the Korea National Health and Nutrition Examination Surveys. The training data set was used to construct models based on popular machine learning algorithms such as support vector machines (SVM), random forests, artificial neural networks (ANN), and logistic regression (LR) based on simple surveys. The machine learning models were compared to four conventional clinical decision tools: osteoporosis self-assessment tool (OST), osteoporosis risk assessment instrument (ORAI), simple calculated osteoporosis risk estimation (SCORE), and osteoporosis index of risk (OSIRIS). RESULTS: SVM had significantly better area under the curve (AUC) of the receiver operating characteristic than ANN, LR, OST, ORAI, SCORE, and OSIRIS for the training set. SVM predicted osteoporosis risk with an AUC of 0.827, accuracy of 76.7%, sensitivity of 77.8%, and specificity of 76.0% at total hip, femoral neck, or lumbar spine for the testing set. The significant factors selected by SVM were age, height, weight, body mass index, duration of menopause, duration of breast feeding, estrogen therapy, hyperlipidemia, hypertension, osteoarthritis, and diabetes mellitus. CONCLUSION: Considering various predictors associated with low bone density, the machine learning methods may be effective tools for identifying postmenopausal women at high risk for osteoporosis.

Osteoporosis risk prediction using machine learning and conventional methods.

A number of clinical decision tools for osteoporosis risk assessment have been developed to select postmenopausal women for the measurement of bone mineral density. We developed and validated machine learning models with the aim of more accurately identifying the risk of osteoporosis in postmenopausal women, and compared with the ability of a conventional clinical decision tool, osteoporosis self-assessment tool (OST). We collected medical records from Korean postmenopausal women based on the Korea National Health and Nutrition Surveys (KNHANES V-1). The training data set was used to construct models based on popular machine learning algorithms such as support vector machines (SVM), random forests (RF), artificial neural networks (ANN), and logistic regression (LR) based on various predictors associated with low bone density. The learning models were compared with OST. SVM had significantly better area under the curve (AUC) of the receiver operating characteristic (ROC) than ANN, LR, and OST. Validation on the test set showed that SVM predicted osteoporosis risk with an AUC of 0.827, accuracy of 76.7%, sensitivity of 77.8%, and specificity of 76.0%. We were the first to perform comparisons of the performance of osteoporosis prediction between the machine learning and conventional methods using population-based epidemiological data. The machine learning methods may be effective tools for identifying postmenopausal women at high risk for osteoporosis.

Interpretation of movement during stair ascent for predicting severity and prognosis of knee osteoarthritis in elderly women using support vector machine.

Several studies have demonstrated that pathologic movement changes in knee osteoarthritis (OA) may contribute to disease progression. The aim of this study was to investigate the association between movement changes during stair ascent and pain, radiographic severity, and prognosis of knee OA in the elderly women using machine learning (ML) over a seven-year follow-up period. Eighteen elderly female patients with knee OA and 20 healthy controls were enrolled. Kinematic data for stair ascent were obtained using a 3D-motion analysis system at baseline. Kinematic factors were analyzed based on one of the popular ML methods, support vector machines (SVM). SVM was used to search kinematic predictors associated with pain, radiographic severity of knee OA, and unfavorable outcomes, which were defined as persistent knee pain as reported at the seven-year follow-up or as having undergone total knee replacement during the follow-up period. Six patients (46.2%) had unfavorable outcomes at the seven-year follow-up. SVM showed accuracy of detection of knee OA (97.4%), prediction of pain (83.3%), radiographic severity (83.3%), and unfavorable outcomes (69.2%). The predictors with SVM included the time of stair ascent, maximal anterior pelvis tilting, knee flexion at initial foot contact, and ankle dorsiflexion at initial foot contact. The interpretation of movement during stair ascent using ML may be helpful for physicians not only in detecting knee OA, but also in evaluating pain and radiographic severity.

Mortality prediction of rats in acute hemorrhagic shock using machine learning techniques.

This study sought to determine a mortality prediction model that could be used for triage in the setting of acute hemorrhage from trauma. To achieve this aim, various machine learning techniques were applied using the rat model in acute hemorrhage. Thirty-six anesthetized rats were randomized into three groups according to the volume of controlled blood loss. Measurements included heart rate (HR), systolic and diastolic blood pressures (SBP and DBP), mean arterial pressure, pulse pressure, respiratory rate, temperature, blood lactate concentration (LC), peripheral perfusion (PP), shock index (SI, SI = HR/SBP), and a new hemorrhage-induced severity index (NI, NI = LC/PP). NI was suggested as one of the good candidates for mortality prediction variable in our previous study. We constructed mortality prediction models with logistic regression (LR), artificial neural networks (ANN), random forest (RF), and support vector machines (SVM) with variable selection. The SVM model showed better sensitivity (1.000) and area under curve (0.972) than the LR, ANN, and RF models for mortality prediction. The important variables selected by the SVM were NI and LC. The SVM model may be very helpful to first responders who need to make accurate triage decisions and rapidly treat hemorrhagic patients in cases of trauma.

Effects of 60 Hz magnetic fields on teenagers and adults.

BACKGROUND: As use of electrical devices has increased, social concerns about the possible effects of 60 Hz electromagnetic fields on human health have increased. Accordingly, the number of people who complain of various symptoms such as headache and insomnia has risen. Many previous studies of the effects of extremely low frequency (ELF) magnetic field exposure on children have focused on the occurrence of childhood leukaemia and central nervous system cancers. However, very few provocation studies have examined the health effects of ELF magnetic fields on teenagers. METHODS: In this double-blind study, we simultaneously investigated physiological changes (heart rate, respiration rate, and heart rate variability), subjective symptoms, and magnetic field perception to determine the reliable effects of 60 Hz 12.5 µT magnetic fields on teenagers. Two volunteer groups of 30 adults and 30 teenagers were tested with exposure to sham and real magnetic fields for 32 min. RESULTS: ELF magnetic field exposure did not have any effects on the physiological parameters or eight subjective symptoms in either group. Neither group correctly perceived the magnetic fields. CONCLUSIONS: Physiological data were analysed, subjective symptoms surveyed, and the percentages of those who believed they were being exposed were measured. No effects were observed in adults or teenagers resulting from 32 min of 60 Hz 12.5 µT magnetic field exposure.

Effects of radiation emitted by WCDMA mobile phones on electromagnetic hypersensitive subjects.

BACKGROUND: With the use of the third generation (3 G) mobile phones on the rise, social concerns have arisen concerning the possible health effects of radio frequency-electromagnetic fields (RF-EMFs) emitted by wideband code division multiple access (WCDMA) mobile phones in humans. The number of people with self-reported electromagnetic hypersensitivity (EHS), who complain of various subjective symptoms such as headache, dizziness and fatigue, has also increased. However, the origins of EHS remain unclear. METHODS: In this double-blind study, two volunteer groups of 17 EHS and 20 non-EHS subjects were simultaneously investigated for physiological changes (heart rate, heart rate variability, and respiration rate), eight subjective symptoms, and perception of RF-EMFs during real and sham exposure sessions. Experiments were conducted using a dummy phone containing a WCDMA module (average power, 24 dBm at 1950 MHz; specific absorption rate, 1.57 W/kg) within a headset placed on the head for 32 min. RESULTS: WCDMA RF-EMFs generated no physiological changes or subjective symptoms in either group. There was no evidence that EHS subjects perceived RF-EMFs better than non-EHS subjects. CONCLUSIONS: Considering the analyzed physiological data, the subjective symptoms surveyed, and the percentages of those who believed they were being exposed, 32 min of RF radiation emitted by WCDMA mobile phones demonstrated no effects in either EHS or non-EHS subjects.

EHS subjects do not perceive RF EMF emitted from smart phones better than non-EHS subjects.

As the use of smart phones increases, social concerns have arisen concerning the possible effects of radio frequency-electromagnetic fields (RF-EMFs) emitted from wideband code division multiple access (WCDMA) mobile phones on human health. The number of people with self-reported electromagnetic hypersensitivity (EHS) who complain of various subjective symptoms, such as headache, insomnia, etc., has also recently increased. However, it is unclear whether EHS subjects can detect RF-EMFs exposure or not. In this double-blind study, two volunteer groups of 17 EHS and 20 non-EHS subjects were investigated in regards to their perception of RF-EMFs with real and sham exposure sessions. Experiments were conducted using a WCDMA module inside a dummy phone with an average power of 24 dBm at 1950 MHz and a specific absorption rate of 1.57 W/kg using a dummy headphone for 32 min. In conclusion, there was no indication that EHS subjects perceive RF-EMFs better than non-EHS subjects.

A survival prediction model of rats in hemorrhagic shock using the random forest classifier.

Hemorrhagic shock is the cause of one third of deaths resulting from injury in the world. Although many studies have tried to diagnose hemorrhagic shock early and accurately, such attempts were inconclusive due to compensatory mechanisms of humans. The objective of this study was to construct a survival prediction model of rats in hemorrhagic shock using a random forest (RF) model, which is a newly emerged classifier acknowledged for its performance. Heart rate (HR), mean arterial pressure (MAP), respiratory rate (RR), lactate concentration (LC), and perfusion (PF) measured in rats were used as input variables for the RF model and its performance was compared with that of a logistic regression (LR) model. Before constructing the models, we performed a 5-fold cross validation for RF variable selection and forward stepwise variable selection for the LR model to see which variables are important for the models. For the LR model, sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (ROC-AUC) were 1, 0.89, 0.94, and 0.98, respectively. For the RF models, sensitivity, specificity, accuracy, and AUC were 0.96, 1, 0.98, and 0.99, respectively. In conclusion, the RF model was superior to the LR model for survival prediction in the rat model.