AI-based analysis of CT images for rapid triage of COVID-19 patients

Utilizing machine learning to predict hospital admissions for pediatric COVID-19 patients (PrepCOVID-Machine)

The COVID-19 pandemic has burdened healthcare systems globally. To curb high hospital admission rates, only patients with genuine medical needs are admitted. However, machine learning (ML) models to predict COVID-19 hospitalization in Asian children are lacking. This study aimed to develop and validate ML models to predict pediatric COVID-19 hospitalization. We collected secondary data with 2200 patients and 65 variables from Malaysian aged 0 to 12 with COVID-19 between 1st February 2020 and 31st March 2022. The sample was partitioned into training, internal, and external validation groups. Recursive Feature Elimination (RFE) was employed for feature selection, and we trained seven supervised classifiers. Grid Search was used to optimize the hyperparameters of each algorithm. The study analyzed 1988 children and 30 study variables after data were processed. The RFE algorithm selected 12 highly predicted variables for COVID-19 hospitalization, including age, male sex, fever, cough, rhinorrhea, shortness of breath, vomiting, diarrhea, seizures, body temperature, chest indrawing, and abnormal breath sounds. With external validation, Adaptive Boosting was the highest-performing classifier (AUROC = 0.95) to predict COVID-19 hospital admission in children. We validated AdaBoost as the best to predict COVID-19 hospitalization among children. This model may assist front-line clinicians in making medical disposition decisions.

A tri-light warning system for hospitalized COVID-19 patients: Credibility-based risk stratification for future pandemic preparedness

Purpose

The novel coronavirus pneumonia (COVID-19) has continually spread and mutated, requiring a patient risk stratification system to optimize medical resources and improve pandemic response. We aimed to develop a conformal prediction-based tri-light warning system for stratifying COVID-19 patients, applicable to both original and emerging variants.

Methods

We retrospectively collected data from 3646 patients across multiple centers in China. The dataset was divided into a training set (n = 1451), a validation set (n = 662), an external test set from Huoshenshan Field Hospital (n = 1263), and a specific test set for Delta and Omicron variants (n = 544). The tri-light warning system extracts radiomic features from CT (computed tomography) and integrates clinical records to classify patients into high-risk (red), uncertain-risk (yellow), and low-risk (green) categories. Models were built to predict ICU (intensive care unit) admissions (adverse cases in training/validation/Huoshenshan/variant test sets: n = 39/21/262/11) and were evaluated using AUROC ((area under the receiver operating characteristic curve)) and AUPRC ((area under the precision-recall curve)) metrics.

Results

The dataset included 1830 men (50.2 %) and 1816 women (50.8 %), with a median age of 53.7 years (IQR [interquartile range]: 42-65 years). The system demonstrated strong performance under data distribution shifts, with AUROC of 0.89 and AUPRC of 0.42 for original strains, and AUROC of 0.77-0.85 and AUPRC of 0.51-0.60 for variants.

Conclusion

The tri-light warning system can enhance pandemic responses by effectively stratifying COVID-19 patients under varying conditions and data shifts.

Testing the generalizability and effectiveness of deep learning models among clinics: sperm detection as a pilot study

BACKGROUND

Deep learning has been increasingly investigated for assisting clinical in vitro fertilization (IVF). The first technical step in many tasks is to visually detect and locate sperm, oocytes, and embryos in images. For clinical deployment of such deep learning models, different clinics use different image acquisition hardware and different sample preprocessing protocols, raising the concern over whether the reported accuracy of a deep learning model by one clinic could be reproduced in another clinic. Here we aim to investigate the effect of each imaging factor on the generalizability of object detection models, using sperm analysis as a pilot example.

METHODS

Ablation studies were performed using state-of-the-art models for detecting human sperm to quantitatively assess how model precision (false-positive detection) and recall (missed detection) were affected by imaging magnification, imaging mode, and sample preprocessing protocols. The results led to the hypothesis that the richness of image acquisition conditions in a training dataset deterministically affects model generalizability. The hypothesis was tested by first enriching the training dataset with a wide range of imaging conditions, then validated through internal blind tests on new samples and external multi-center clinical validations.

RESULTS

Ablation experiments revealed that removing subsets of data from the training dataset significantly reduced model precision. Removing raw sample images from the training dataset caused the largest drop in model precision, whereas removing 20x images caused the largest drop in model recall. by incorporating different imaging and sample preprocessing conditions into a rich training dataset, the model achieved an intraclass correlation coefficient (ICC) of 0.97 (95% CI: 0.94-0.99) for precision, and an ICC of 0.97 (95% CI: 0.93-0.99) for recall. Multi-center clinical validation showed no significant differences in model precision or recall across different clinics and applications.

CONCLUSIONS

The results validated the hypothesis that the richness of data in the training dataset is a key factor impacting model generalizability. These findings highlight the importance of diversity in a training dataset for model evaluation and suggest that future deep learning models in andrology and reproductive medicine should incorporate comprehensive feature sets for enhanced generalizability across clinics.

Multimodal data fusion using sparse canonical correlation analysis and cooperative learning: a COVID-19 cohort study

Through technological innovations, patient cohorts can be examined from multiple views with high-dimensional, multiscale biomedical data to classify clinical phenotypes and predict outcomes. Here, we aim to present our approach for analyzing multimodal data using unsupervised and supervised sparse linear methods in a COVID-19 patient cohort. This prospective cohort study of 149 adult patients was conducted in a tertiary care academic center. First, we used sparse canonical correlation analysis (CCA) to identify and quantify relationships across different data modalities, including viral genome sequencing, imaging, clinical data, and laboratory results. Then, we used cooperative learning to predict the clinical outcome of COVID-19 patients: Intensive care unit admission. We show that serum biomarkers representing severe disease and acute phase response correlate with original and wavelet radiomics features in the LLL frequency channel (cor(Xu1, Zv1) = 0.596, p value < 0.001). Among radiomics features, histogram-based first-order features reporting the skewness, kurtosis, and uniformity have the lowest negative, whereas entropy-related features have the highest positive coefficients. Moreover, unsupervised analysis of clinical data and laboratory results gives insights into distinct clinical phenotypes. Leveraging the availability of global viral genome databases, we demonstrate that the Word2Vec natural language processing model can be used for viral genome encoding. It not only separates major SARS-CoV-2 variants but also allows the preservation of phylogenetic relationships among them. Our quadruple model using Word2Vec encoding achieves better prediction results in the supervised task. The model yields area under the curve (AUC) and accuracy values of 0.87 and 0.77, respectively. Our study illustrates that sparse CCA analysis and cooperative learning are powerful techniques for handling high-dimensional, multimodal data to investigate multivariate associations in unsupervised and supervised tasks.

Multimodal Biomedical Data Fusion Using Sparse Canonical Correlation Analysis and Cooperative Learning: A Cohort Study on COVID-19

Through technological innovations, patient cohorts can be examined from multiple views with high-dimensional, multiscale biomedical data to classify clinical phenotypes and predict outcomes. Here, we aim to present our approach for analyzing multimodal data using unsupervised and supervised sparse linear methods in a COVID-19 patient cohort. This prospective cohort study of 149 adult patients was conducted in a tertiary care academic center. First, we used sparse canonical correlation analysis (CCA) to identify and quantify relationships across different data modalities, including viral genome sequencing, imaging, clinical data, and laboratory results. Then, we used cooperative learning to predict the clinical outcome of COVID-19 patients. We show that serum biomarkers representing severe disease and acute phase response correlate with original and wavelet radiomics features in the LLL frequency channel (corr(Xu1, Zv1) = 0.596, p-value < 0.001). Among radiomics features, histogram-based first-order features reporting the skewness, kurtosis, and uniformity have the lowest negative, whereas entropy-related features have the highest positive coefficients. Moreover, unsupervised analysis of clinical data and laboratory results gives insights into distinct clinical phenotypes. Leveraging the availability of global viral genome databases, we demonstrate that the Word2Vec natural language processing model can be used for viral genome encoding. It not only separates major SARS-CoV-2 variants but also allows the preservation of phylogenetic relationships among them. Our quadruple model using Word2Vec encoding achieves better prediction results in the supervised task. The model yields area under the curve (AUC) and accuracy values of 0.87 and 0.77, respectively. Our study illustrates that sparse CCA analysis and cooperative learning are powerful techniques for handling high-dimensional, multimodal data to investigate multivariate associations in unsupervised and supervised tasks.

Machine-learning-based head impact subtyping based on the spectral densities of the measurable head kinematics

BACKGROUND

Traumatic brain injury can be caused by head impacts, but many brain injury risk estimation models are not equally accurate across the variety of impacts that patients may undergo, and the characteristics of different types of impacts are not well studied. We investigated the spectral characteristics of different head impact types with kinematics classification.

METHODS

Data were analyzed from 3262 head impacts from lab reconstruction, American football, mixed martial arts, and publicly available car crash data. A random forest classifier with spectral densities of linear acceleration and angular velocity was built to classify head impact types (e.g., football, car crash, mixed martial arts). To test the classifier robustness, another 271 lab-reconstructed impacts were obtained from 5 other instrumented mouthguards. Finally, with the classifier, type-specific, nearest-neighbor regression models were built for brain strain.

RESULTS

The classifier reached a median accuracy of 96% over 1000 random partitions of training and test sets. The most important features in the classification included both low- and high-frequency features, both linear acceleration features and angular velocity features. Different head impact types had different distributions of spectral densities in low- and high-frequency ranges (e.g., the spectral densities of mixed martial arts impacts were higher in the high-frequency range than in the low-frequency range). The type-specific regression showed a generally higher R2 value than baseline models without classification.

CONCLUSION

The machine-learning-based classifier enables a better understanding of the impact kinematics spectral density in different sports, and it can be applied to evaluate the quality of impact-simulation systems and on-field data augmentation.

A transformer-based representation-learning model with unified processing of multimodal input for clinical diagnostics

During the diagnostic process, clinicians leverage multimodal information, such as the chief complaint, medical images and laboratory test results. Deep-learning models for aiding diagnosis have yet to meet this requirement of leveraging multimodal information. Here we report a transformer-based representation-learning model as a clinical diagnostic aid that processes multimodal input in a unified manner. Rather than learning modality-specific features, the model leverages embedding layers to convert images and unstructured and structured text into visual tokens and text tokens, and uses bidirectional blocks with intramodal and intermodal attention to learn holistic representations of radiographs, the unstructured chief complaint and clinical history, and structured clinical information such as laboratory test results and patient demographic information. The unified model outperformed an image-only model and non-unified multimodal diagnosis models in the identification of pulmonary disease (by 12% and 9%, respectively) and in the prediction of adverse clinical outcomes in patients with COVID-19 (by 29% and 7%, respectively). Unified multimodal transformer-based models may help streamline the triaging of patients and facilitate the clinical decision-making process.

Artificial intelligence-based model for COVID-19 prognosis incorporating chest radiographs and clinical data; a retrospective model development and validation study

OBJECTIVES

The purpose of this study was to develop an artificial intelligence-based model to prognosticate COVID-19 patients at admission by combining clinical data and chest radiographs.

METHODS

This retrospective study used the Stony Brook University COVID-19 dataset of 1384 inpatients. After exclusions, 1356 patients were randomly divided into training (1083) and test datasets (273). We implemented three artificial intelligence models, which classified mortality, ICU admission, or ventilation risk. Each model had three submodels with different inputs: clinical data, chest radiographs, and both. We showed the importance of the variables using SHapley Additive exPlanations (SHAP) values.

RESULTS

The mortality prediction model was best overall with area under the curve, sensitivity, specificity, and accuracy of 0.79 (0.72-0.86), 0.74 (0.68-0.79), 0.77 (0.61-0.88), and 0.74 (0.69-0.79) for the clinical data-based model; 0.77 (0.69-0.85), 0.67 (0.61-0.73), 0.81 (0.67-0.92), 0.70 (0.64-0.75) for the image-based model, and 0.86 (0.81-0.91), 0.76 (0.70-0.81), 0.77 (0.61-0.88), 0.76 (0.70-0.81) for the mixed model. The mixed model had the best performance (p value < 0.05). The radiographs ranked fourth for prognostication overall, and first of the inpatient tests assessed.

CONCLUSIONS

These results suggest that prognosis models become more accurate if AI-derived chest radiograph features and clinical data are used together.

ADVANCES IN KNOWLEDGE

This AI model evaluates chest radiographs together with clinical data in order to classify patients as having high or low mortality risk. This work shows that chest radiographs taken at admission have significant COVID-19 prognostic information compared to clinical data other than age and sex.

Data science in the intensive care unit

In this editorial, we comment on the current development and deployment of data science in intensive care units (ICUs). Data in ICUs can be classified into qualitative and quantitative data with different technologies needed to translate and interpret them. Data science, in the form of artificial intelligence (AI), should find the right interaction between physicians, data and algorithm. For individual patients and physicians, sepsis and mechanical ventilation have been two important aspects where AI has been extensively studied. However, major risks of bias, lack of generalizability and poor clinical values remain. AI deployment in the ICUs should be emphasized more to facilitate AI development. For ICU management, AI has a huge potential in transforming resource allocation. The coronavirus disease 2019 pandemic has given opportunities to establish such systems which should be investigated further. Ethical concerns must be addressed when designing such AI.

Development and validation of an abnormality-derived deep-learning diagnostic system for major respiratory diseases

Respiratory diseases impose a tremendous global health burden on large patient populations. In this study, we aimed to develop DeepMRD^TR, a deep learning-based medical image interpretation system for the diagnosis of major respiratory diseases based on the automated identification of a wide range of radiological abnormalities through computed tomography (CT) and chest X-ray (CXR) from real-world, large-scale datasets. DeepMRD^TR comprises four networks (two CT-Nets and two CXR-Nets) that exploit contrastive learning to generate pre-training parameters that are fine-tuned on the retrospective dataset collected from a single institution. The performance of DeepMRD^TR was evaluated for abnormality identification and disease diagnosis on data from two different institutions: one was an internal testing dataset from the same institution as the training data and the second was collected from an external institution to evaluate the model generalizability and robustness to an unrelated population dataset. In such a difficult multi-class diagnosis task, our system achieved the average area under the receiver operating characteristic curve (AUC) of 0.856 (95% confidence interval (CI):0.843–0.868) and 0.841 (95%CI:0.832–0.887) for abnormality identification, and 0.900 (95%CI:0.872–0.958) and 0.866 (95%CI:0.832–0.887) for major respiratory diseases’ diagnosis on CT and CXR datasets, respectively. Furthermore, to achieve a clinically actionable diagnosis, we deployed a preliminary version of DeepMRD^TR into the clinical workflow, which was performed on par with senior experts in disease diagnosis, with an AUC of 0.890 and a Cohen’s k of 0.746–0.877 at a reasonable timescale; these findings demonstrate the potential to accelerate the medical workflow to facilitate early diagnosis as a triage tool for respiratory diseases which supports improved clinical diagnoses and decision-making.

COVID-19 mortality prediction in the intensive care unit with deep learning based on longitudinal chest X-rays and clinical data

OBJECTIVES

We aimed to develop deep learning models using longitudinal chest X-rays (CXRs) and clinical data to predict in-hospital mortality of COVID-19 patients in the intensive care unit (ICU).

METHODS

Six hundred fifty-four patients (212 deceased, 442 alive, 5645 total CXRs) were identified across two institutions. Imaging and clinical data from one institution were used to train five longitudinal transformer-based networks applying five-fold cross-validation. The models were tested on data from the other institution, and pairwise comparisons were used to determine the best-performing models.

RESULTS

A higher proportion of deceased patients had elevated white blood cell count, decreased absolute lymphocyte count, elevated creatine concentration, and incidence of cardiovascular and chronic kidney disease. A model based on pre-ICU CXRs achieved an AUC of 0.632 and an accuracy of 0.593, and a model based on ICU CXRs achieved an AUC of 0.697 and an accuracy of 0.657. A model based on all longitudinal CXRs (both pre-ICU and ICU) achieved an AUC of 0.702 and an accuracy of 0.694. A model based on clinical data alone achieved an AUC of 0.653 and an accuracy of 0.657. The addition of longitudinal imaging to clinical data in a combined model significantly improved performance, reaching an AUC of 0.727 (p = 0.039) and an accuracy of 0.732.

CONCLUSIONS

The addition of longitudinal CXRs to clinical data significantly improves mortality prediction with deep learning for COVID-19 patients in the ICU.

KEY POINTS

• Deep learning was used to predict mortality in COVID-19 ICU patients. • Serial radiographs and clinical data were used. • The models could inform clinical decision-making and resource allocation.

Machine learning decision tree algorithm role for predicting mortality in critically ill adult COVID-19 patients admitted to the ICU

Background

Coronavirus disease-19 (COVID-19) is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and is currently a major cause of intensive care unit (ICU) admissions globally. The role of machine learning in the ICU is evolving but currently limited to diagnostic and prognostic values. A decision tree (DT) algorithm is a simple and intuitive machine learning method that provides sequential nonlinear analysis of variables. It is simple and might be a valuable tool for bedside physicians during COVID-19 to predict ICU outcomes and help in critical decision-making like end-of-life decisions and bed allocation in the event of limited ICU bed capacities. Herein, we utilized a machine learning DT algorithm to describe the association of a predefined set of variables and 28-day ICU outcome in adult COVID-19 patients admitted to the ICU. We highlight the value of utilizing a machine learning DT algorithm in the ICU at the time of a COVID-19 pandemic.

Methods

This was a prospective and multicenter cohort study involving 14 hospitals in Saudi Arabia. We included critically ill COVID-19 patients admitted to the ICU between March 1, 2020, and October 31, 2020. The predictors of 28-day ICU mortality were identified using two predictive models: conventional logistic regression and DT analyses.

Results

There were 1468 critically ill COVID-19 patients included in the study. The 28-day ICU mortality was 540 (36.8 %), and the 90-day mortality was 600 (40.9 %). The DT algorithm identified five variables that were integrated into the algorithm to predict 28-day ICU outcomes: need for intubation, need for vasopressors, age, gender, and PaO2/FiO2 ratio.

Conclusion

DT is a simple tool that might be utilized in the ICU to identify critically ill COVID-19 patients who are at high risk of 28-day ICU mortality. However, further studies and external validation are still required.

A Brief Analysis of a New Device to Prevent Early Intubation in Hypoxemic Patients: An Observational Study

The need for mechanical ventilation is one of the main concerns related to the care of patients with COVID-19. The aim of this study is to evaluate the efficacy of a bubble device for oxygen supplementation. This device was implemented for the selected patients hospitalized with severe COVID-19 pneumonia with persistent low oxygen saturation. Patients were selected in three major COVID-19 hospitals of Bahia state in Brazil from July to November 2020, where they remained with the device for seven days and were monitored for different factors, such as vital signs, oximetry evaluation, and arterial blood gasometry. Among the 51 patients included in the study, 68.63% successfully overcame hypoxemia without the necessity to be transferred to mechanical ventilation, whereas 31.37% required tracheal intubation (p value < 0.05). There was no difference of note on the analysis of the clinical data, chemistry, and hematological evaluation, with the exception of the SpO2 on follow-up days. Multivariate analysis revealed that the independent variable, male sex, SpO2, and non-inhaled mask, was associated with the necessity of requiring early mechanical ventilation. We concluded that this bubble device should be a prior step to be utilized before indication of mechanical ventilation in patients with persistent hypoxemia of severe COVID-19 pneumonia.

Deep learning-based lesion subtyping and prediction of clinical outcomes in COVID-19 pneumonia using chest CT

The main objective of this work is to develop and evaluate an artificial intelligence system based on deep learning capable of automatically identifying, quantifying, and characterizing COVID-19 pneumonia patterns in order to assess disease severity and predict clinical outcomes, and to compare the prediction performance with respect to human reader severity assessment and whole lung radiomics. We propose a deep learning based scheme to automatically segment the different lesion subtypes in nonenhanced CT scans. The automatic lesion quantification was used to predict clinical outcomes. The proposed technique has been independently tested in a multicentric cohort of 103 patients, retrospectively collected between March and July of 2020. Segmentation of lesion subtypes was evaluated using both overlapping (Dice) and distance-based (Hausdorff and average surface) metrics, while the proposed system to predict clinically relevant outcomes was assessed using the area under the curve (AUC). Additionally, other metrics including sensitivity, specificity, positive predictive value and negative predictive value were estimated. 95% confidence intervals were properly calculated. The agreement between the automatic estimate of parenchymal damage (%) and the radiologists' severity scoring was strong, with a Spearman correlation coefficient (R) of 0.83. The automatic quantification of lesion subtypes was able to predict patient mortality, admission to the Intensive Care Units (ICU) and need for mechanical ventilation with an AUC of 0.87, 0.73 and 0.68 respectively. The proposed artificial intelligence system enabled a better prediction of those clinically relevant outcomes when compared to the radiologists' interpretation and to whole lung radiomics. In conclusion, deep learning lesion subtyping in COVID-19 pneumonia from noncontrast chest CT enables quantitative assessment of disease severity and better prediction of clinical outcomes with respect to whole lung radiomics or radiologists' severity score.

Sounds of COVID-19: exploring realistic performance of audio-based digital testing

To identify Coronavirus disease (COVID-19) cases efficiently, affordably, and at scale, recent work has shown how audio (including cough, breathing and voice) based approaches can be used for testing. However, there is a lack of exploration of how biases and methodological decisions impact these tools' performance in practice. In this paper, we explore the realistic performance of audio-based digital testing of COVID-19. To investigate this, we collected a large crowdsourced respiratory audio dataset through a mobile app, alongside symptoms and COVID-19 test results. Within the collected dataset, we selected 5240 samples from 2478 English-speaking participants and split them into participant-independent sets for model development and validation. In addition to controlling the language, we also balanced demographics for model training to avoid potential acoustic bias. We used these audio samples to construct an audio-based COVID-19 prediction model. The unbiased model took features extracted from breathing, coughs and voice signals as predictors and yielded an AUC-ROC of 0.71 (95% CI: 0.65-0.77). We further explored several scenarios with different types of unbalanced data distributions to demonstrate how biases and participant splits affect the performance. With these different, but less appropriate, evaluation strategies, the performance could be overestimated, reaching an AUC up to 0.90 (95% CI: 0.85-0.95) in some circumstances. We found that an unrealistic experimental setting can result in misleading, sometimes over-optimistic, performance. Instead, we reported complete and reliable results on crowd-sourced data, which would allow medical professionals and policy makers to accurately assess the value of this technology and facilitate its deployment.

A catalogue of tools and variables from crisis and routine care to support decision-making about allocation of intensive care beds and ventilator treatment during pandemics: Scoping review

PURPOSE

This scoping review sought to identify objective factors to assist clinicians and policy-makers in making consistent, objective and ethically sound decisions about resource allocation when healthcare rationing is inevitable.

MATERIALS AND METHODS

Review of guidelines and tools used in ICUs, hospital wards and emergency departments on how to best allocate intensive care beds and ventilators either during routine care or developed during previous epidemics, and association with patient outcomes during and after hospitalisation.

RESULTS

Eighty publications from 20 countries reporting accuracy or validity of prognostic tools/algorithms, or significant correlation between prognostic variables and clinical outcomes met our eligibility criteria: twelve pandemic guidelines/triage protocols/consensus statements, twenty-two pandemic algorithms, and 46 prognostic tools/variables from non-crisis situations. Prognostic indicators presented here can be combined to create locally-relevant triage algorithms for clinicians and policy makers deciding about allocation of ICU beds and ventilators during a pandemic. No consensus was found on the ethical issues to incorporate in the decision to admit or triage out of intensive care.

CONCLUSIONS

This review provides a unique reference intended as a discussion starter for clinicians and policy makers to consider formalising an objective a locally-relevant triage consensus document that enhances confidence in decision-making during healthcare rationing of critical care and ventilator resources.

Role of chest imaging in the diagnosis and treatment of COVID-19

Background: Thousands of new patients are diagnosed with coronavirus disease 2019 (COVID-19) daily worldwide. We reviewed the role of chest imaging in the diagnosis and treatment of patients with COVID-19.Current Concepts: Chest imaging is not recommended as a primary diagnostic tool for COVID-19. However, when real-time polymerase chain reaction is difficult to perform or when COVID-19 is strongly suspected, chest imaging can assist in the diagnosis. Thus, chest imaging is recommended for high-risk patients and patients with worsening respiratory symptoms, but not for asymptomatic patients. Bilateral peripheral pneumonia is a typical imaging finding in patients with COVID-19. However, there are cases where chest imaging shows atypical findings or appears normal. The extent of COVID-19 pneumonia on chest imaging is related to the severity of the disease. The presence and extent of pneumonia on chest imaging can help monitor patients, select appropriate treatment agents, determine whether the patient should be hospitalized, and predict the prognosis.Discussion and Conclusion: Appropriate use of chest imaging is needed for clinicians to help triage patients with COVID-19 and decide on the treatment plan.

Structuring clinical text with AI: Old versus new natural language processing techniques evaluated on eight common cardiovascular diseases

Free-text clinical notes in electronic health records are more difficult for data mining while the structured diagnostic codes can be missing or erroneous. To improve the quality of diagnostic codes, this work extracts diagnostic codes from free-text notes: five old and new word vectorization methods were used to vectorize Stanford progress notes and predict eight ICD-10 codes of common cardiovascular diseases with logistic regression. The models showed good performance, with TF-IDF as the best vectorization model showing the highest AUROC (0.9499-0.9915) and AUPRC (0.2956-0.8072). The models also showed transferability when tested on MIMIC-III data with AUROC from 0.7952 to 0.9790 and AUPRC from 0.2353 to 0.8084. Model interpretability was shown by the important words with clinical meanings matching each disease. This study shows the feasibility of accurately extracting structured diagnostic codes, imputing missing codes, and correcting erroneous codes from free-text clinical notes for information retrieval and downstream machine-learning applications.