Explainable Machine Learning Model for Glaucoma Diagnosis and Its Interpretation.
Ontology highlight
ABSTRACT:
SUBMITTER: Oh S
PROVIDER: S-EPMC8001225 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Background and objectivesAccurate predictions of haemoglobin (Hb) deferral for whole-blood donors could aid blood banks in reducing deferral rates and increasing efficiency and donor motivation. Complex models are needed to make accurate predictions, but predictions must also be explainable. Before the implementation of a prediction model, its impact on the blood supply should be estimated to avoid shortages.Materials and methodsDonation visits between October 2017 and December 2021 were selected from Sanquin's database system. The following variables were available for each visit: donor sex, age, donation start time, month, number of donations in the last 24 months, most recent ferritin level, days since last ferritin measurement, Hb at nth previous visit (n between 1 and 5), days since the nth previous visit. Outcome Hb deferral has two classes: deferred and not deferred. Support vector machines were used as prediction models, and SHapley Additive exPlanations values were used to quantify the contribution of each variable to the model predictions. Performance was assessed using precision and recall. The potential impact on blood supply was estimated by predicting deferral at earlier or later donation dates.ResultsWe present a model that predicts Hb deferral in an explainable way. If used in practice, 64% of non-deferred donors would be invited on or before their original donation date, while 80% of deferred donors would be invited later.ConclusionBy using this model to invite donors, the number of blood bank visits would increase by 15%, while deferral rates would decrease by 60% (currently 3% for women and 1% for men).
| S-EPMC9826045 | biostudies-literature
Project description:Many challenges persist in developing accurate computational models for predicting solvation free energy (ΔGsol). Despite recent developments in Machine Learning (ML) methodologies that outperformed traditional quantum mechanical models, several issues remain concerning explanatory insights for broad chemical predictions with an acceptable speed-accuracy trade-off. To overcome this, we present a novel supervised ML model to predict the ΔGsol for an array of solvent-solute pairs. Using two different ensemble regressor algorithms, we made fast and accurate property predictions using open-source chemical features, encoding complex electronic, structural, and surface area descriptors for every solvent and solute. By integrating molecular properties and chemical interaction features, we have analyzed individual descriptor importance and optimized our model though explanatory information form feature groups. On aqueous and organic solvent databases, ML models revealed the predictive relevance of solutes with increasing polar surface area and decreasing polarizability, yielding better results than state-of-the-art benchmark Neural Network methods (without complex quantum mechanical or molecular dynamic simulations). Both algorithms successfully outperformed previous ΔGsol predictions methods, with a maximum absolute error of 0.22 ± 0.02 kcal mol-1, further validated in an external benchmark database and with solvent hold-out tests. With these explanatory and statistical insights, they allow a thoughtful application of this method for predicting other thermodynamic properties, stressing the relevance of ML modeling for further complex computational chemistry problems.
| S-EPMC11005042 | biostudies-literature
Project description:ObjectivesThe aim of this study is to determine whether the clinical features including blood markers can establish an explainable machine learning model to predict epidermal growth factor receptor (EGFR) mutation in lung cancer.MethodsWe retrospectively analyzed 7,413 patients with lung adenocarcinoma (LA) diagnosed by gene sequencing in West China Hospital of the Sichuan University from April 2015 to June 2019. The machine learning algorithms (MLAs) included logistic regression (LR), random forest (RF), LightGBM, support vector machine (SVM), multi-layer perceptron (MLP), extreme gradient boosting (XGBoost), and decision tree (DT). Demographic characteristics, personal history, and blood markers were taken into. The area under the receiver operating characteristic curve (AUC) and SHapley Additive exPlanation (SHAP) value were used to explain the prediction models.ResultsOf the 7,413 patients with LA (47.6%), 3,527 were identified with EGFR mutation; RF achieved greatest performance in predicting EGFR mutation AUC [0.771, 95% confidence interval (CI): 0.770, 0.772], which was like XGBoost with AUC (0.740, 95% CI: 0.739, 0.741). The five most influential features were smoking consumption, sex, cholesterol, age, and albumin globulin ratio. The SHAP summary and dependence plot have been used to explain the affection of the 12 features to this model and how a single feature influences the output, respectively.ConclusionWe established EGFR mutation prediction models by MLAs and revealed that the RF was preferred, AUC (0.771, 95% CI: 0.770, 0.772), which was better than the traditional models. Therefore, the artificial intelligence-based MLA predicting model may become a practical tool to guide in diagnosis and therapy of LA.
| S-EPMC9259982 | biostudies-literature
Project description:IntroductionTraditional Alzheimer's disease (AD) and mild cognitive impairment (MCI) screening lacks the sensitivity and timeliness required to detect subtle indicators of cognitive decline. Multimodal artificial intelligence technologies using only speech data promise improved detection of neurodegenerative disorders.MethodsSpeech collected over the telephone from 91 older participants who were cognitively healthy (n = 29) or had diagnoses of AD (n = 30) or amnestic MCI (aMCI; n = 32) was analyzed with multimodal natural language and speech processing methods. An explainable ensemble decision tree classifier for the multiclass prediction of cognitive decline was created.ResultsThis approach was 75% accurate overall-an improvement over traditional speech-based screening tools and a unimodal language-based model. We include a dashboard for the examination of the results, allowing for novel ways of interpreting such data.DiscussionThis work provides a foundation for a meaningful change in medicine as clinical translation, scalability, and user friendliness were core to the methodologies.HighlightsRemote assessments and artificial intelligence (AI) models allow greater access to cognitive decline screening.Speech impairments differ significantly between mild AD, amnestic mild cognitive impairment (aMCI), and healthy controls.AI predictions of cognitive decline are more accurate than experts and standard tools.The AI model was 75% accurate in classifying mild AD, aMCI, and healthy controls.
| S-EPMC10752754 | biostudies-literature
Project description:Early identification of high-risk metabolic dysfunction-associated steatohepatitis (MASH) can offer patients access to novel therapeutic options and potentially decrease the risk of progression to cirrhosis. This study aimed to develop an explainable machine learning model for high-risk MASH prediction and compare its performance with well-established biomarkers. Data were derived from the National Health and Nutrition Examination Surveys (NHANES) 2017-March 2020, which included a total of 5281 adults with valid elastography measurements. We used a FAST score ≥ 0.35, calculated using liver stiffness measurement and controlled attenuation parameter values and aspartate aminotransferase levels, to identify individuals with high-risk MASH. We developed an ensemble-based machine learning XGBoost model to detect high-risk MASH and explored the model's interpretability using an explainable artificial intelligence SHAP method. The prevalence of high-risk MASH was 6.9%. Our XGBoost model achieved a high level of sensitivity (0.82), specificity (0.91), accuracy (0.90), and AUC (0.95) for identifying high-risk MASH. Our model demonstrated a superior ability to predict high-risk MASH vs. FIB-4, APRI, BARD, and MASLD fibrosis scores (AUC of 0.95 vs. 0.50, 0.50, 0.49 and 0.50, respectively). To explain the high performance of our model, we found that the top 5 predictors of high-risk MASH were ALT, GGT, platelet count, waist circumference, and age. We used an explainable ML approach to develop a clinically applicable model that outperforms commonly used clinical risk indices and could increase the identification of high-risk MASH patients in resource-limited settings.
| S-EPMC11016071 | biostudies-literature
Project description:Spontaneous bacterial peritonitis (SBP) is a life-threatening complication in patients with cirrhosis. We aimed to develop an explainable machine learning model to achieve the early prediction and outcome interpretation of SBP. We used CatBoost algorithm to construct MODEL-1 with 46 variables. After dimensionality reduction, we constructed MODEL-2. We calculated and compared the sensitivity and negative predictive value (NPV) of MODEL-1 and MODEL-2. Finally, we used the SHAP (SHapley Additive exPlanations) method to provide insights into the model's outcome or prediction. MODEL-2 (AUROC: 0.822; 95% confidence interval [CI] 0.783-0.856), liked MODEL-1 (AUROC: 0.822; 95% CI 0.784-0.856), could well predict the risk of SBP in cirrhotic ascites patients. The 6 most influential predictive variables were total protein, C-reactive protein, prothrombin activity, cholinesterase, lymphocyte ratio and apolipoprotein A1. For binary classifier, the sensitivity and NPV of MODEL-1 were 0.894 and 0.885, respectively, while for MODEL-2 they were 0.927 and 0.904, respectively. We applied CatBoost algorithm to establish a practical and explainable prediction model for risk of SBP in cirrhotic patients with ascites. We also identified 6 important variables closely related to the occurrence of SBP.
| S-EPMC8569162 | biostudies-literature
Project description:Electroencephalography (EEG) is a non-invasive method employed to discern human behaviors by monitoring the neurological responses during cognitive and motor tasks. Machine learning (ML) represents a promising tool for the recognition of human activities (HAR), and eXplainable artificial intelligence (XAI) can elucidate the role of EEG features in ML-based HAR models. The primary objective of this investigation is to investigate the feasibility of an EEG-based ML model for categorizing everyday activities, such as resting, motor, and cognitive tasks, and interpreting models clinically through XAI techniques to explicate the EEG features that contribute the most to different HAR states. The study involved an examination of 75 healthy individuals with no prior diagnosis of neurological disorders. EEG recordings were obtained during the resting state, as well as two motor control states (walking and working tasks), and a cognition state (reading task). Electrodes were placed in specific regions of the brain, including the frontal, central, temporal, and occipital lobes (Fz, C1, C2, T7, T8, Oz). Several ML models were trained using EEG data for activity recognition and LIME (Local Interpretable Model-Agnostic Explanations) was employed for interpreting clinically the most influential EEG spectral features in HAR models. The classification results of the HAR models, particularly the Random Forest and Gradient Boosting models, demonstrated outstanding performances in distinguishing the analyzed human activities. The ML models exhibited alignment with EEG spectral bands in the recognition of human activity, a finding supported by the XAI explanations. To sum up, incorporating eXplainable Artificial Intelligence (XAI) into Human Activity Recognition (HAR) studies may improve activity monitoring for patient recovery, motor imagery, the healthcare metaverse, and clinical virtual reality settings.
| S-EPMC10490625 | biostudies-literature
Project description:BackgroundWe aimed to develop and validate an automated machine learning (autoML) prediction model for cardiac surgery-associated acute kidney injury (CSA-AKI).MethodsUsing 69 preoperative variables, we developed several models to predict post-operative AKI in adult patients undergoing cardiac surgery. Models included autoML and non-autoML types, including decision tree (DT), random forest (RF), extreme gradient boosting (XGBoost), and artificial neural network (ANN), as well as a logistic regression prediction model. We then compared model performance using area under the receiver operating characteristic curve (AUROC) and assessed model calibration using Brier score on the independent testing dataset.ResultsThe incidence of CSA-AKI was 36%. Stacked ensemble autoML had the highest predictive performance among autoML models, and was chosen for comparison with other non-autoML and multivariable logistic regression models. The autoML had the highest AUROC (0.79), followed by RF (0.78), XGBoost (0.77), multivariable logistic regression (0.77), ANN (0.75), and DT (0.64). The autoML had comparable AUROC with RF and outperformed the other models. The autoML was well-calibrated. The Brier score for autoML, RF, DT, XGBoost, ANN, and multivariable logistic regression was 0.18, 0.18, 0.21, 0.19, 0.19, and 0.18, respectively. We applied SHAP and LIME algorithms to our autoML prediction model to extract an explanation of the variables that drive patient-specific predictions of CSA-AKI.ConclusionWe were able to present a preoperative autoML prediction model for CSA-AKI that provided high predictive performance that was comparable to RF and superior to other ML and multivariable logistic regression models. The novel approaches of the proposed explainable preoperative autoML prediction model for CSA-AKI may guide clinicians in advancing individualized medicine plans for patients under cardiac surgery.
| S-EPMC9656700 | biostudies-literature
Project description:Early detection is important in glaucoma management. By using optical coherence tomography (OCT), the subtle structural changes caused by glaucoma can be detected. Though OCT provided abundant parameters for comprehensive information, clinicians may be confused once the results conflict. Machine learning classifiers (MLCs) are good tools for considering numerous parameters and generating reliable diagnoses in glaucoma practice. Here we aim to compare different MLCs based on Spectralis OCT parameters, including circumpapillary retinal nerve fiber layer (cRNFL) thickness, Bruch's membrane opening-minimum rim width (BMO-MRW), Early Treatment Diabetes Retinopathy Study (ETDRS) macular thickness, and posterior pole asymmetry analysis (PPAA), in discriminating normal from glaucomatous eyes. Five MLCs were proposed, namely conditional inference trees (CIT), logistic model tree (LMT), C5.0 decision tree, random forest (RF), and extreme gradient boosting (XGBoost). Logistic regression (LGR) was used as a benchmark for comparison. RF was shown to be the best model. Ganglion cell layer measurements were the most important predictors in early glaucoma detection and cRNFL measurements were more important as the glaucoma severity increased. The global, temporal, inferior, superotemporal, and inferotemporal sites were relatively influential locations among all parameters. Clinicians should cautiously integrate the Spectralis OCT results into the entire clinical picture when diagnosing glaucoma.
| S-EPMC8471622 | biostudies-literature
Project description:Alzheimer's disease (AD) has its onset many decades before dementia develops, and work is ongoing to characterise individuals at risk of decline on the basis of early detection through biomarker and cognitive testing as well as the presence/absence of identified risk factors. Risk prediction models for AD based on various computational approaches, including machine learning, are being developed with promising results. However, these approaches have been criticised as they are unable to generalise due to over-reliance on one data source, poor internal and external validations, and lack of understanding of prediction models, thereby limiting the clinical utility of these prediction models. We propose a framework that employs a transfer-learning paradigm with ensemble learning algorithms to develop explainable personalised risk prediction models for dementia. Our prediction models, known as source models, are initially trained and tested using a publicly available dataset (n = 84,856, mean age = 69 years) with 14 years of follow-up samples to predict the individual risk of developing dementia. The decision boundaries of the best source model are further updated by using an alternative dataset from a different and much younger population (n = 473, mean age = 52 years) to obtain an additional prediction model known as the target model. We further apply the SHapely Additive exPlanation (SHAP) algorithm to visualise the risk factors responsible for the prediction at both population and individual levels. The best source model achieves a geometric accuracy of 87%, specificity of 99%, and sensitivity of 76%. In comparison to a baseline model, our target model achieves better performance across several performance metrics, within an increase in geometric accuracy of 16.9%, specificity of 2.7%, and sensitivity of 19.1%, an area under the receiver operating curve (AUROC) of 11% and a transfer learning efficacy rate of 20.6%. The strength of our approach is the large sample size used in training the source model, transferring and applying the "knowledge" to another dataset from a different and undiagnosed population for the early detection and prediction of dementia risk, and the ability to visualise the interaction of the risk factors that drive the prediction. This approach has direct clinical utility.
| S-EPMC8187875 | biostudies-literature