Cox Proportional Hazard Regression Versus a Deep Learning Algorithm in the Prediction of Dementia: An Analysis Based on Periodic Health Examination

Deep neural network base competing risk in predicting heart failure patient's survival

Objectives

Heart failure (HF) is a complicated disease with several competing risks of interest, such as HF death and other causes. This study compares a Deep Neural Network Competing Risks (DNNCR) and Random Survival Forest (RSF) model to evaluate the predictive performance of time-to-event outcomes in HF patients with competing risks.

Methods

This study represents the retrospective analysis of 435 HF patients admitted to RCMRC, Tehran, Iran, between March and August 2018. After a five-year follow-up in 2023, predictions were analyzed based on Cause of death. This study employed RSF and DNN methods to account for competing risks in survival analysis. Then, model fitness was applied using C-index and IBS.

Results

The C-index of the results shows that DNNCR is superior to RSF in predicting survival outcomes for HF and other causes of death. Precisely, the C-index was 0.65 (0.04) for HF and 0.63 (0.02) for other causes of death in the DNNCR model, while in RSF, the C-index was 0.65 (0.04) for HF and 0.61 (0.03) for Other Causes. Additionally, calibration results showed via the IBS the finest performance of the DNNCR model at 0.16 for HF, followed by other causes with an IBS of 0.18.

Conclusions

The study shows that the DNNCR model outperforms RSF in predicting survival outcomes for HF patients, particularly in the presence of competing risks. The improved accuracy enables physicians to identify high-risk individuals and tailor treatment plans accordingly. Future research could utilize more diverse datasets to enhance DNNCR performance and integrate these models into clinical tools.

Graphical Abstract

Precision fetal cardiology detects cyanotic congenital heart disease using maternal saliva metabolome and artificial intelligence

Prenatal sonographic diagnosis of congenital heart disease (CHD) can lead to improved morbidity and mortality. However, the diagnostic accuracy of ultrasound, the sole prenatal screening tool, remains limited. Failed prenatal or early newborn detection of cyanotic CHD (CCHD) can have disastrous consequences. We therefore sought to use a Precision Fetal Cardiology based approach combining metabolomic profiling of maternal saliva and machine learning, a major branch of artificial intelligence (AI), for the prenatal detection of isolated, non-syndromic cyanotic CHD. Metabolomic analyses using Ultra-High Performance Liquid Chromatography/Mass Spectrometry identified 468 metabolites in the saliva. Six different AI platforms were utilized for the detection of CCHD and CHD overall. AI achieved excellent accuracy for the CCHD detection: Area Under the ROC curve: AUC (95% CI) = 0.819 (0.635-1.00) with a sensitivity and specificity of 92.5% and 87.0%, and for CHD overall: AUC (95% CI) = 0.828 (0.635-1.00) with a sensitivity of 90.5% and specificity of 88.0%. Similarly high accuracies were achieved for the detection of CHD overall: AUC (95% CI) = 0.8488 (0.635-1.00) with a sensitivity of 92.5% and specificity of 91.0%. Pathway analysis showed significant alterations in Arachidonic Acid, Alpha-linoleic acid, and Tryptophan metabolism indicating significant lipid dysfunction in cyanotic CHD. In summary, we report for the first time, the accurate detection of non-syndromic cyanotic CHD using maternal salivary metabolomics. Further, analysis revealed significant alteration of lipid metabolism.

Early Identification of Cognitive Impairment in Community Environments Through Modeling Subtle Inconsistencies in Questionnaire Responses: Machine Learning Model Development and Validation

BACKGROUND

The underdiagnosis of cognitive impairment hinders timely intervention of dementia. Health professionals working in the community play a critical role in the early detection of cognitive impairment, yet still face several challenges such as a lack of suitable tools, necessary training, and potential stigmatization.

OBJECTIVE

This study explored a novel application integrating psychometric methods with data science techniques to model subtle inconsistencies in questionnaire response data for early identification of cognitive impairment in community environments.

METHODS

This study analyzed questionnaire response data from participants aged 50 years and older in the Health and Retirement Study (waves 8-9, n=12,942). Predictors included low-quality response indices generated using the graded response model from four brief questionnaires (optimism, hopelessness, purpose in life, and life satisfaction) assessing aspects of overall well-being, a focus of health professionals in communities. The primary and supplemental predicted outcomes were current cognitive impairment derived from a validated criterion and dementia or mortality in the next ten years. Seven predictive models were trained, and the performance of these models was evaluated and compared.

RESULTS

The multilayer perceptron exhibited the best performance in predicting current cognitive impairment. In the selected four questionnaires, the area under curve values for identifying current cognitive impairment ranged from 0.63 to 0.66 and was improved to 0.71 to 0.74 when combining the low-quality response indices with age and gender for prediction. We set the threshold for assessing cognitive impairment risk in the tool based on the ratio of underdiagnosis costs to overdiagnosis costs, and a ratio of 4 was used as the default choice. Furthermore, the tool outperformed the efficiency of age or health-based screening strategies for identifying individuals at high risk for cognitive impairment, particularly in the 50- to 59-year and 60- to 69-year age groups. The tool is available on a portal website for the public to access freely.

CONCLUSIONS

We developed a novel prediction tool that integrates psychometric methods with data science to facilitate "passive or backend" cognitive impairment assessments in community settings, aiming to promote early cognitive impairment detection. This tool simplifies the cognitive impairment assessment process, making it more adaptable and reducing burdens. Our approach also presents a new perspective for using questionnaire data: leveraging, rather than dismissing, low-quality data.

A deep learning model for Alzheimer's disease diagnosis based on patient clinical records

BACKGROUND

Dementia, with Alzheimer's disease (AD) being the most common type of this neurodegenerative disease, is an under-diagnosed health problem in older people. The creation of classification models based on AD risk factors using Deep Learning is a promising tool to minimize the impact of under-diagnosis.

OBJECTIVE

To develop a Deep Learning model that uses clinical data from patients with dementia to classify whether they have AD.

METHODS

A Deep Learning model to identify AD in clinical records is proposed. In addition, several rebalancing methods have been used to preprocess the dataset and several studies have been carried out to tune up the model.

RESULTS

Model has been tested against other well-established machine learning techniques, having better results than these in terms of AUC with alpha less than 0.05.

CONCLUSIONS

The developed Neural Network Model has a good performance and can be an accurate assisting tool for AD diagnosis.

The combination of olfactory dysfunction and depression increases the risk of incident dementia in older adults

OBJECTIVES

Olfactory dysfunction and depression are common in later life, and both have been presented as risk factors for dementia. Our purpose was to investigate the associations between these two risk factors and determine if they had an additive effect on dementia risk.

DESIGN

Olfactory function was assessed using the Brief Smell Identification Test (BSIT), and depression was classified using a combination of the 15-item Geriatric Depression Scale (GDS) score and current antidepressant use. Cross-sectional associations between depression and olfactory function were examined using correlations. Cox regression analyses were conducted to examine the longitudinal relationship between olfaction and depression and incident dementia across 12-years of follow-up.

PARTICIPANTS

Participants were 780 older adults (aged 70-90 years; 56.5% female) from the Sydney Memory and Ageing Study (MAS) without a diagnosis of dementia at baseline.

RESULTS

Partial correlation revealed a nonsignificant association between baseline depression and olfactory function after accounting for covariates (r = -.051, p = .173). Cox regression showed that depression at baseline (hazard ratio = 1.706, 95% CI 1.185-2.456, p = .004) and lower BSIT scores (HR = .845, 95%CI .789-.905, p < .001) were independently associated with a higher risk of incident dementia across 12 years. Entering both predictors together improved the overall predictive power of the model.

CONCLUSIONS

Lower olfactory identification scores and depressive symptoms predict incident dementia over 12 years. The use of BSIT scores and depression in conjunction provides a greater ability to predict dementia than either used alone. Assessment of olfactory function and depression screening may provide clinical utility in the early detection of dementia.

What's New in Dementia Risk Prediction Modelling? An Updated Systematic Review

Introduction

Identifying individuals at high risk of dementia is critical to optimized clinical care, formulating effective preventative strategies, and determining eligibility for clinical trials. Since our previous systematic reviews in 2010 and 2015, there has been a surge in dementia risk prediction modelling. The aim of this study was to update our previous reviews to explore, and critically review, new developments in dementia risk modelling.

Methods

MEDLINE, Embase, Scopus, and Web of Science were searched from March 2014 to June 2022. Studies were included if they were population- or community-based cohorts (including electronic health record data), had developed a model for predicting late-life incident dementia, and included model performance indices such as discrimination, calibration, or external validation.

Results

In total, 9,209 articles were identified from the electronic search, of which 74 met the inclusion criteria. We found a substantial increase in the number of new models published from 2014 (>50 new models), including an increase in the number of models developed using machine learning. Over 450 unique predictor (component) variables have been tested. Nineteen studies (26%) undertook external validation of newly developed or existing models, with mixed results. For the first time, models have also been developed in low- and middle-income countries (LMICs) and others validated in racial and ethnic minority groups.

Conclusion

The literature on dementia risk prediction modelling is rapidly evolving with new analytical developments and testing in LMICs. However, it is still challenging to make recommendations about which one model is the most suitable for routine use in a clinical setting. There is an urgent need to develop a suitable, robust, validated risk prediction model in the general population that can be widely implemented in clinical practice to improve dementia prevention.

Neural Computation-Based Methods for the Early Diagnosis and Prognosis of Alzheimer's Disease Not Using Neuroimaging Biomarkers: A Systematic Review

Background

The growing number of older adults in recent decades has led to more prevalent geriatric diseases, such as strokes and dementia. Therefore, Alzheimer's disease (AD), as the most common type of dementia, has become more frequent too.

Background

Objective: The goals of this work are to present state-of-the-art studies focused on the automatic diagnosis and prognosis of AD and its early stages, mainly mild cognitive impairment, and predicting how the research on this topic may change in the future.

Methods

Articles found in the existing literature needed to fulfill several selection criteria. Among others, their classification methods were based on artificial neural networks (ANNs), including deep learning, and data not from brain signals or neuroimaging techniques were used. Considering our selection criteria, 42 articles published in the last decade were finally selected.

Results

The most medically significant results are shown. Similar quantities of articles based on shallow and deep ANNs were found. Recurrent neural networks and transformers were common with speech or in longitudinal studies. Convolutional neural networks (CNNs) were popular with gait or combined with others in modular approaches. Above one third of the cross-sectional studies utilized multimodal data. Non-public datasets were frequently used in cross-sectional studies, whereas the opposite in longitudinal ones. The most popular databases were indicated, which will be helpful for future researchers in this field.

Conclusions

The introduction of CNNs in the last decade and their superb results with neuroimaging data did not negatively affect the usage of other modalities. In fact, new ones emerged.

Precision oncology: Artificial intelligence, circulating cell-free DNA, and the minimally invasive detection of pancreatic cancer-A pilot study

BACKGROUND

Pancreatic cancer (PC) is among the most lethal cancers. The lack of effective tools for early detection results in late tumor detection and, consequently, high mortality rate. Precision oncology aims to develop targeted individual treatments based on advanced computational approaches of omics data. Biomarkers, such as global alteration of cytosine (CpG) methylation, can be pivotal for these objectives. In this study, we performed DNA methylation profiling of pancreatic cancer patients using circulating cell-free DNA (cfDNA) and artificial intelligence (AI) including Deep Learning (DL) for minimally invasive detection to elucidate the epigenetic pathogenesis of PC.

METHODS

The Illumina Infinium HD Assay was used for genome-wide DNA methylation profiling of cfDNA in treatment-naïve patients. Six AI algorithms were used to determine PC detection accuracy based on cytosine (CpG) methylation markers. Additional strategies for minimizing overfitting were employed. The molecular pathogenesis was interrogated using enrichment analysis.

RESULTS

In total, we identified 4556 significantly differentially methylated CpGs (q-value < 0.05; Bonferroni correction) in PC versus controls. Highly accurate PC detection was achieved with all 6 AI platforms (Area under the receiver operator characteristics curve [0.90-1.00]). For example, DL achieved AUC (95% CI): 1.00 (0.95-1.00), with a sensitivity and specificity of 100%. A separate modeling approach based on logistic regression-based yielded an AUC (95% CI) 1.0 (1.0-1.0) with a sensitivity and specificity of 100% for PC detection. The top four biological pathways that were epigenetically altered in PC and are known to be linked with cancer are discussed.

CONCLUSION

Using a minimally invasive approach, AI, and epigenetic analysis of circulating cfDNA, high predictive accuracy for PC was achieved. From a clinical perspective, our findings suggest that that early detection leading to improved overall survival may be achievable in the future.

Detection of Cancer Recurrence Using Systemic Inflammatory Markers and Machine Learning after Concurrent Chemoradiotherapy for Head and Neck Cancers

Pretreatment values of the neutrophil-to-lymphocyte ratio (NLR) and the platelet-to-lymphocyte ratio (PLR) are well-established prognosticators in various cancers, including head and neck cancers. However, there are no studies on whether temporal changes in the NLR and PLR values after treatment are related to the development of recurrence. Therefore, in this study, we aimed to develop a deep neural network (DNN) model to discern cancer recurrence from temporal NLR and PLR values during follow-up after concurrent chemoradiotherapy (CCRT) and to evaluate the model's performance compared with conventional machine learning (ML) models. Along with conventional ML models such as logistic regression (LR), random forest (RF), and gradient boosting (GB), the DNN model to discern recurrences was trained using a dataset of 778 consecutive patients with primary head and neck cancers who received CCRT. There were 16 input features used, including 12 laboratory values related to the NLR and the PLR. Along with the original training dataset (N = 778), data were augmented to split the training dataset (N = 900). The model performance was measured using ROC-AUC and PR-AUC values. External validation was performed using a dataset of 173 patients from an unrelated external institution. The ROC-AUC and PR-AUC values of the DNN model were 0.828 ± 0.032 and 0.663 ± 0.069, respectively, in the original training dataset, which were higher than the ROC-AUC and PR-AUC values of the LR, RF, and GB models in the original training dataset. With the recursive feature elimination (RFE) algorithm, five input features were selected. The ROC-AUC and PR-AUC values of the DNN-RFE model were higher than those of the original DNN model (0.883 ± 0.027 and 0.778 ± 0.042, respectively). The ROC-AUC and PR-AUC values of the DNN-RFE model trained with a split dataset were 0.889 ± 0.032 and 0.771 ± 0.044, respectively. In the external validation, the ROC-AUC values of the DNN-RFE model trained with the original dataset and the same model trained with the split dataset were 0.710 and 0.784, respectively. The DNN model with feature selection using the RFE algorithm showed the best performance among the ML models to discern a recurrence after CCRT in patients with head and neck cancers. Data augmentation by splitting training data was helpful for model performance. The performance of the DNN-RFE model was also validated with an external dataset.

Multi-domain prognostic models used in middle-aged adults without known cognitive impairment for predicting subsequent dementia

BACKGROUND

Dementia, a global health priority, has no current cure. Around 50 million people worldwide currently live with dementia, and this number is expected to treble by 2050. Some health conditions and lifestyle behaviours can increase or decrease the risk of dementia and are known as 'predictors'. Prognostic models combine such predictors to measure the risk of future dementia. Models that can accurately predict future dementia would help clinicians select high-risk adults in middle age and implement targeted risk reduction.

OBJECTIVES

Our primary objective was to identify multi-domain prognostic models used in middle-aged adults (aged 45 to 65 years) for predicting dementia or cognitive impairment. Eligible multi-domain prognostic models involved two or more of the modifiable dementia predictors identified in a 2020 Lancet Commission report and a 2019 World Health Organization (WHO) report (less education, hearing loss, traumatic brain injury, hypertension, excessive alcohol intake, obesity, smoking, depression, social isolation, physical inactivity, diabetes mellitus, air pollution, poor diet, and cognitive inactivity). Our secondary objectives were to summarise the prognostic models, to appraise their predictive accuracy (discrimination and calibration) as reported in the development and validation studies, and to identify the implications of using dementia prognostic models for the management of people at a higher risk for future dementia.

SEARCH METHODS

We searched MEDLINE, Embase, PsycINFO, CINAHL, and ISI Web of Science Core Collection from inception until 6 June 2022. We performed forwards and backwards citation tracking of included studies using the Web of Science platform.  SELECTION CRITERIA: We included development and validation studies of multi-domain prognostic models. The minimum eligible follow-up was five years. Our primary outcome was an incident clinical diagnosis of dementia based on validated diagnostic criteria, and our secondary outcome was dementia or cognitive impairment determined by any other method.

DATA COLLECTION AND ANALYSIS

Two review authors independently screened the references, extracted data using a template based on the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS), and assessed risk of bias and applicability of included studies using the Prediction model Risk Of Bias ASsessment Tool (PROBAST). We synthesised the C-statistics of models that had been externally validated in at least three comparable studies.  MAIN RESULTS: We identified 20 eligible studies; eight were development studies and 12 were validation studies. There were 14 unique prognostic models: seven models with validation studies and seven models with development-only studies. The models included a median of nine predictors (range 6 to 34); the median number of modifiable predictors was five (range 2 to 11). The most common modifiable predictors in externally validated models were diabetes, hypertension, smoking, physical activity, and obesity. In development-only models, the most common modifiable predictors were obesity, diabetes, hypertension, and smoking. No models included hearing loss or air pollution as predictors. Nineteen studies had a high risk of bias according to the PROBAST assessment, mainly because of inappropriate analysis methods, particularly lack of reported calibration measures. Applicability concerns were low for 12 studies, as their population, predictors, and outcomes were consistent with those of interest for this review. Applicability concerns were high for nine studies, as they lacked baseline cognitive screening or excluded an age group within the range of 45 to 65 years. Only one model, Cardiovascular Risk Factors, Ageing, and Dementia (CAIDE), had been externally validated in multiple studies, allowing for meta-analysis. The CAIDE model included eight predictors (four modifiable predictors): age, education, sex, systolic blood pressure, body mass index (BMI), total cholesterol, physical activity and APOEƐ4 status. Overall, our confidence in the prediction accuracy of CAIDE was very low; our main reasons for downgrading the certainty of the evidence were high risk of bias across all the studies, high concern of applicability, non-overlapping confidence intervals (CIs), and a high degree of heterogeneity. The summary C-statistic was 0.71 (95% CI 0.66 to 0.76; 3 studies; very low-certainty evidence) for the incident clinical diagnosis of dementia, and 0.67 (95% CI 0.61 to 0.73; 3 studies; very low-certainty evidence) for dementia or cognitive impairment based on cognitive scores. Meta-analysis of calibration measures was not possible, as few studies provided these data.

AUTHORS' CONCLUSIONS

We identified 14 unique multi-domain prognostic models used in middle-aged adults for predicting subsequent dementia. Diabetes, hypertension, obesity, and smoking were the most common modifiable risk factors used as predictors in the models. We performed meta-analyses of C-statistics for one model (CAIDE), but the summary values were unreliable. Owing to lack of data, we were unable to meta-analyse the calibration measures of CAIDE. This review highlights the need for further robust external validations of multi-domain prognostic models for predicting future risk of dementia in middle-aged adults.

Machine learning for predicting neurodegenerative diseases in the general older population: a cohort study

BACKGROUND

In the older general population, neurodegenerative diseases (NDs) are associated with increased disability, decreased physical and cognitive function. Detecting risk factors can help implement prevention measures. Using deep neural networks (DNNs), a machine-learning algorithm could be an alternative to Cox regression in tabular datasets with many predictive features. We aimed to compare the performance of different types of DNNs with regularized Cox proportional hazards models to predict NDs in the older general population.

METHODS

We performed a longitudinal analysis with participants of the English Longitudinal Study of Ageing. We included men and women with no NDs at baseline, aged 60 years and older, assessed every 2 years from 2004 to 2005 (wave2) to 2016-2017 (wave 8). The features were a set of 91 epidemiological and clinical baseline variables. The outcome was new events of Parkinson's, Alzheimer or dementia. After applying multiple imputations, we trained three DNN algorithms: Feedforward, TabTransformer, and Dense Convolutional (Densenet). In addition, we trained two algorithms based on Cox models: Elastic Net regularization (CoxEn) and selected features (CoxSf).

RESULTS

5433 participants were included in wave 2. During follow-up, 12.7% participants developed NDs. Although the five models predicted NDs events, the discriminative ability was superior using TabTransformer (Uno's C-statistic (coefficient (95% confidence intervals)) 0.757 (0.702, 0.805). TabTransformer showed superior time-dependent balanced accuracy (0.834 (0.779, 0.889)) and specificity (0.855 (0.0.773, 0.909)) than the other models. With the CoxSf (hazard ratio (95% confidence intervals)), age (10.0 (6.9, 14.7)), poor hearing (1.3 (1.1, 1.5)) and weight loss 1.3 (1.1, 1.6)) were associated with a higher DNN risk. In contrast, executive function (0.3 (0.2, 0.6)), memory (0, 0, 0.1)), increased gait speed (0.2, (0.1, 0.4)), vigorous physical activity (0.7, 0.6, 0.9)) and higher BMI (0.4 (0.2, 0.8)) were associated with a lower DNN risk.

CONCLUSION

TabTransformer is promising for prediction of NDs with heterogeneous tabular datasets with numerous features. Moreover, it can handle censored data. However, Cox models perform well and are easier to interpret than DNNs. Therefore, they are still a good choice for NDs.

Using machine learning algorithms for predicting cognitive impairment and identifying modifiable factors among Chinese elderly people

Objectives: This study firstly aimed to explore predicting cognitive impairment at an early stage using a large population-based longitudinal survey of elderly Chinese people. The second aim was to identify reversible factors which may help slow the rate of decline in cognitive function over 3 years in the community. Methods: We included 12,280 elderly people from four waves of the Chinese Longitudinal Healthy Longevity Survey (CLHLS), followed from 2002 to 2014. The Chinese version of the Mini-Mental State Examination (MMSE) was used to examine cognitive function. Six machine learning algorithms (including a neural network model) and an ensemble method were trained on data split 2/3 for training and 1/3 testing. Parameters were explored in training data using 3-fold cross-validation and models were evaluated in test data. The model performance was measured by area-under-curve (AUC), sensitivity, and specificity. In addition, due to its better interpretability, logistic regression (LR) was used to assess the association of life behavior and its change with cognitive impairment after 3 years. Results: Support vector machine and multi-layer perceptron were found to be the best performing algorithms with AUC of 0.8267 and 0.8256, respectively. Fusing the results of all six single models further improves the AUC to 0.8269. Playing more Mahjong or cards (OR = 0.49,95% CI: 0.38-0.64), doing more garden works (OR = 0.54,95% CI: 0.43-0.68), watching TV or listening to the radio more (OR = 0.67,95% CI: 0.59-0.77) were associated with decreased risk of cognitive impairment after 3 years. Conclusions: Machine learning algorithms especially the SVM, and the ensemble model can be leveraged to identify the elderly at risk of cognitive impairment. Doing more leisure activities, doing more gardening work, and engaging in more activities combined were associated with decreased risk of cognitive impairment.

Determining a minimum set of variables for machine learning cardiovascular event prediction: results from REFINE SPECT registry

AIMS

Optimal risk stratification with machine learning (ML) from myocardial perfusion imaging (MPI) includes both clinical and imaging data. While most imaging variables can be derived automatically, clinical variables require manual collection, which is time-consuming and prone to error. We determined the fewest manually input and imaging variables required to maintain the prognostic accuracy for major adverse cardiac events (MACE) in patients undergoing a single-photon emission computed tomography (SPECT) MPI.

METHODS AND RESULTS

This study included 20 414 patients from the multicentre REFINE SPECT registry and 2984 from the University of Calgary for training and external testing of the ML models, respectively. ML models were trained using all variables (ML-All) and all image-derived variables (including age and sex, ML-Image). Next, ML models were sequentially trained by incrementally adding manually input and imaging variables to baseline ML models based on their importance ranking. The fewest variables were determined as the ML models (ML-Reduced, ML-Minimum, and ML-Image-Reduced) that achieved comparable prognostic performance to ML-All and ML-Image. Prognostic accuracy of the ML models was compared with visual diagnosis, stress total perfusion deficit (TPD), and traditional multivariable models using area under the receiver-operating characteristic curve (AUC). ML-Minimum (AUC 0.798) obtained comparable prognostic accuracy to ML-All (AUC 0.799, P = 0.19) by including 12 of 40 manually input variables and 11 of 58 imaging variables. ML-Reduced achieved comparable accuracy (AUC 0.796) with a reduced set of manually input variables and all imaging variables. In external validation, the ML models also obtained comparable or higher prognostic accuracy than traditional multivariable models.

CONCLUSION

Reduced ML models, including a minimum set of manually collected or imaging variables, achieved slightly lower accuracy compared to a full ML model but outperformed standard interpretation methods and risk models. ML models with fewer collected variables may be more practical for clinical implementation.

Deep-learning model for predicting the survival of rectal adenocarcinoma patients based on a surveillance, epidemiology, and end results analysis

BACKGROUND

We collected information on patients with rectal adenocarcinoma in the United States from the Surveillance, Epidemiology, and EndResults (SEER) database. We used this information to establish a model that combined deep learning with a multilayer neural network (the DeepSurv model) for predicting the survival rate of patients with rectal adenocarcinoma.

METHODS

We collected patients with rectal adenocarcinoma in the United States and older than 20 yearswho had been added to the SEER database from 2004 to 2015. We divided these patients into training and test cohortsat a ratio of 7:3. The training cohort was used to develop a seven-layer neural network based on the analysis method established by Katzman and colleagues to construct a DeepSurv prediction model. We then used the C-index and calibration plots to evaluate the prediction performance of the DeepSurv model.

RESULTS

The 49,275 patients with rectal adenocarcinoma included in the study were randomly divided into the training cohort (70%, n = 34,492) and the test cohort (30%, n = 14,783). There were no statistically significant differences in clinical characteristics between the two cohorts (p > 0.05). We applied Cox proportional-hazards regression to the data in the training cohort, which showed that age, sex, marital status, tumor grade, surgery status, and chemotherapy status were significant factors influencing survival (p < 0.05). Using the training cohort to construct the DeepSurv model resulted in a C-index of the model of 0.824, while using the test cohort to verify the DeepSurv model yielded a C-index of 0.821. Thesevalues show that the prediction effect of the DeepSurv model for the test-cohort patients was highly consistent with the prediction resultsfor the training-cohort patients.

CONCLUSION

The DeepSurv prediction model of the seven-layer neural network that we have established can accurately predict the survival rateand time of rectal adenocarcinoma patients.

Deep neural network models for identifying incident dementia using claims and EHR datasets

This study investigates the use of deep learning methods to improve the accuracy of a predictive model for dementia, and compares the performance to a traditional machine learning model. With sufficient accuracy the model can be deployed as a first round screening tool for clinical follow-up including neurological examination, neuropsychological testing, imaging and recruitment to clinical trials. Seven cohorts with two years of data, three to eight years prior to index date, and an incident cohort were created. Four trained models for each cohort, boosted trees, feed forward network, recurrent neural network and recurrent neural network with pre-trained weights, were constructed and their performance compared using validation and test data. The incident model had an AUC of 94.4% and F1 score of 54.1%. Eight years removed from index date the AUC and F1 scores were 80.7% and 25.6%, respectively. The results for the remaining cohorts were between these ranges. Deep learning models can result in significant improvement in performance but come at a cost in terms of run times and hardware requirements. The results of the model at index date indicate that this modeling can be effective at stratifying patients at risk of dementia. At this time, the inability to sustain this quality at longer lead times is more an issue of data availability and quality rather than one of algorithm choices.