Cross-Domain Transfer Learning for PCG Diagnosis Algorithm

A time series driven model for early sepsis prediction based on transformer module

Sepsis remains a critical concern in intensive care units due to its high mortality rate. Early identification and intervention are paramount to improving patient outcomes. In this study, we have proposed predictive models for early sepsis prediction based on time-series data, utilizing both CNN-Transformer and LSTM-Transformer architectures. By collecting time-series data from patients at 4, 8, and 12 h prior to sepsis diagnosis and subjecting it to various network models for analysis and comparison. In contrast to traditional recurrent neural networks, our model exhibited a substantial improvement of approximately 20%. On average, our model demonstrated an accuracy of 0.964 (± 0.018), a precision of 0.956 (± 0.012), a recall of 0.967 (± 0.012), and an F1 score of 0.959 (± 0.014). Furthermore, by adjusting the time window, it was observed that the Transformer-based model demonstrated exceptional predictive capabilities, particularly within the earlier time window (i.e., 12 h before onset), thus holding significant promise for early clinical diagnosis and intervention. Besides, we employed the SHAP algorithm to visualize the weight distribution of different features, enhancing the interpretability of our model and facilitating early clinical diagnosis and intervention.

Early-warning of Cardiac Condition through Detection of Murmur in Heart Sound - A Case Study

Cardiovascular disease, particularly Rheumatic Heart Disease (RHD), is one of the leading causes of death in many developing countries. RHD is manageable and treatable with early detection. However, multiple countries across the globe suffer from a scarcity of experienced physicians who can perform screening at large scales. Advancements in machine learning and signal processing have paved way for Phonocardiogram (PCG)-based automatic heart sound classification. The direct implication of such methods is that it is possible to enable a person without specialized training to detect potential cardiac conditions with just a digital stethoscope. Hospitalization or life-threatening situations can be dramatically reduced via such early screenings. Towards this, we conducted a case study amongst a population from a particular geography using machine learning and deep learning methods for the detection of murmur in heart sounds. The methodology consists of first pre-processing and identifying normal vs. abnormal heart sound signals using 3 state-of-the-art methods. The second step further identifies the murmur to be systolic or diastolic by capturing the auscultation location. Abnormal findings are then sent for early attention of clinicians for proper diagnosis. The case study investigates the efficacy of the automated method employed for early screening of potential RHD and initial encouraging results of the study are presented.

An efficient and robust Phonocardiography (PCG)-based Valvular Heart Diseases (VHD) detection framework using Vision Transformer (ViT)

BACKGROUND AND OBJECTIVES

Valvular heart diseases (VHDs) are one of the dominant causes of cardiovascular abnormalities that have been associated with high mortality rates globally. Rapid and accurate diagnosis of the early stage of VHD based on cardiac phonocardiogram (PCG) signal is critical that allows for optimum medication and reduction of mortality rate.

METHODS

To this end, the current study proposes novel deep learning (DL)-based high-performance VHD detection frameworks that are relatively simpler in terms of network structures, yet effective for accurately detecting multiple VHDs. We present three different frameworks considering both 1D and 2D PCG raw signals. For 1D PCG, Mel frequency cepstral coefficients (MFCC) and linear prediction cepstral coefficients (LPCC) features, whereas, for 2D PCG, various deep convolutional neural networks (D-CNNs) features are extracted. Additionally, nature/bio-inspired algorithms (NIA/BIA) including particle swarm optimization (PSO) and genetic algorithm (GA) have been utilized for automatic and efficient feature selection directly from the raw PCG signal. To further improve the performance of the classifier, vision transformer (ViT) has been implemented levering the self-attention mechanism on the time frequency representation (TFR) of 2D PCG signal. Our extensive study presents a comparative performance analysis and the scope of enhancement for the combination of different descriptors, classifiers, and feature selection algorithms.

MAIN RESULTS

Among all classifiers, ViT provides the best performance by achieving mean average accuracy A_cc of 99.90 % and F1-score of 99.95 % outperforming current state-of-the-art VHD classification models.

CONCLUSIONS

The present research provides a robust and efficient DL-based end-to-end PCG signal classification framework for designing a automated high-performance VHD diagnosis system.

A Computer-Aided Heart Valve Disease Diagnosis System Based on Machine Learning

Cardiac auscultation is a noninvasive, convenient, and low-cost diagnostic method for heart valvular disease, and it can diagnose the abnormality of the heart valve at an early stage. However, the accuracy of auscultation relies on the professionalism of cardiologists. Doctors in remote areas may lack the experience to diagnose correctly. Therefore, it is necessary to design a system to assist with the diagnosis. This study proposed a computer-aided heart valve disease diagnosis system, including a heart sound acquisition module, a trained model for diagnosis, and software, which can diagnose four kinds of heart valve diseases. In this study, a training dataset containing five categories of heart sounds was collected, including normal, mitral stenosis, mitral regurgitation, and aortic stenosis heart sound. A convolutional neural network GoogLeNet and weighted KNN are used to train the models separately. For the model trained by the convolutional neural network, time series heart sound signals are converted into time-frequency scalograms based on continuous wavelet transform to adapt to the architecture of GoogLeNet. For the model trained by weighted KNN, features from the time domain and time-frequency domain are extracted manually. Then feature selection based on the chi-square test is performed to get a better group of features. Moreover, we designed software that lets doctors upload heart sounds, visualize the heart sound waveform, and use the model to get the diagnosis. Model assessments using accuracy, sensitivity, specificity, and F1 score indicators are done on two trained models. The results showed that the model trained by modified GoogLeNet outperformed others, with an overall accuracy of 97.5%. The average accuracy, sensitivity, specificity, and F1 score for diagnosing four kinds of heart valve diseases are 98.75%, 96.88%, 99.22%, and 97.99%, respectively. The computer-aided diagnosis system, with a heart sound acquisition module, a diagnostic model, and software, can visualize the heart sound waveform and show the reference diagnostic results. This can assist in the diagnosis of heart valve diseases, especially in remote areas, which lack skilled doctors.

Transferability and interpretability of the sepsis prediction models in the intensive care unit

BACKGROUND

We aimed to develop an early warning system for real-time sepsis prediction in the ICU by machine learning methods, with tools for interpretative analysis of the predictions. In particular, we focus on the deployment of the system in a target medical center with small historical samples.

METHODS

Light Gradient Boosting Machine (LightGBM) and multilayer perceptron (MLP) were trained on Medical Information Mart for Intensive Care (MIMIC-III) dataset and then finetuned on the private Historical Database of local Ruijin Hospital (HDRJH) using transfer learning technique. The Shapley Additive Explanations (SHAP) analysis was employed to characterize the feature importance in the prediction inference. Ultimately, the performance of the sepsis prediction system was further evaluated in the real-world study in the ICU of the target Ruijin Hospital.

RESULTS

The datasets comprised 6891 patients from MIMIC-III, 453 from HDRJH, and 67 from Ruijin real-world data. The area under the receiver operating characteristic curves (AUCs) for LightGBM and MLP models derived from MIMIC-III were 0.98 - 0.98 and 0.95 - 0.96 respectively on MIMIC-III dataset, and, in comparison, 0.82 - 0.86 and 0.84 - 0.87 respectively on HDRJH, from 1 to  5 h preceding. After transfer learning and ensemble learning, the AUCs of the final ensemble model were enhanced to 0.94 - 0.94 on HDRJH and to 0.86 - 0.9 in the real-world study in the ICU of the target Ruijin Hospital. In addition, the SHAP analysis illustrated the importance of age, antibiotics, net balance, and ventilation for sepsis prediction, making the model interpretable.

CONCLUSIONS

Our machine learning model allows accurate real-time prediction of sepsis within 5-h preceding. Transfer learning can effectively improve the feasibility to deploy the prediction model in the target cohort, and ameliorate the model performance for external validation. SHAP analysis indicates that the role of antibiotic usage and fluid management needs further investigation. We argue that our system and methodology have the potential to improve ICU management by helping medical practitioners identify at-sepsis-risk patients and prepare for timely diagnosis and intervention.

TRIAL REGISTRATION

NCT05088850 (retrospectively registered).

An audio processing pipeline for acquiring diagnostic quality heart sounds via mobile phone

Recently, heart sound signals captured using mobile phones have been employed to develop data-driven heart disease detection systems. Such signals are generally captured in person by trained clinicians who can determine if the recorded heart sounds are of diagnosable quality. However, mobile phones have the potential to support heart health diagnostics, even where access to trained medical professionals is limited. To adopt mobile phones as self-diagnostic tools for the masses, we would need to have a mechanism to automatically establish that heart sounds recorded by non-expert users in uncontrolled conditions have the required quality for diagnostic purposes. This paper proposes a quality assessment and enhancement pipeline for heart sounds captured using mobile phones. The pipeline analyzes a heart sound and determines if it has the required quality for diagnostic tasks. Also, in cases where the quality of the captured signal is below the required threshold, the pipeline can improve the quality by applying quality enhancement algorithms. Using this pipeline, we can also provide feedback to users regarding the cause of low-quality signal capture and guide them towards a successful one. We conducted a survey of a group of thirteen clinicians with auscultation skills and experience. The results of this survey were used to inform and validate the proposed quality assessment and enhancement pipeline. We observed a high level of agreement between the survey results and fundamental design decisions within the proposed pipeline. Also, the results indicate that the proposed pipeline can reduce our dependency on trained clinicians for capture of diagnosable heart sounds.

Transfer learning for non-image data in clinical research: A scoping review

BACKGROUND

Transfer learning is a form of machine learning where a pre-trained model trained on a specific task is reused as a starting point and tailored to another task in a different dataset. While transfer learning has garnered considerable attention in medical image analysis, its use for clinical non-image data is not well studied. Therefore, the objective of this scoping review was to explore the use of transfer learning for non-image data in the clinical literature.

METHODS AND FINDINGS

We systematically searched medical databases (PubMed, EMBASE, CINAHL) for peer-reviewed clinical studies that used transfer learning on human non-image data. We included 83 studies in the review. More than half of the studies (63%) were published within 12 months of the search. Transfer learning was most often applied to time series data (61%), followed by tabular data (18%), audio (12%) and text (8%). Thirty-three (40%) studies applied an image-based model to non-image data after transforming data into images (e.g. spectrograms). Twenty-nine (35%) studies did not have any authors with a health-related affiliation. Many studies used publicly available datasets (66%) and models (49%), but fewer shared their code (27%).

CONCLUSIONS

In this scoping review, we have described current trends in the use of transfer learning for non-image data in the clinical literature. We found that the use of transfer learning has grown rapidly within the last few years. We have identified studies and demonstrated the potential of transfer learning in clinical research in a wide range of medical specialties. More interdisciplinary collaborations and the wider adaption of reproducible research principles are needed to increase the impact of transfer learning in clinical research.