A case-based ensemble learning system for explainable breast cancer recurrence prediction

Performance assessment of hybrid machine learning approaches for breast cancer and recurrence prediction

Breast cancer is a major health concern for women everywhere and a major killer of women. Malignant tumors may be distinguished from benign ones, allowing for early diagnosis of this disease. Therefore, doctors need an accurate method of diagnosing tumors as either malignant or benign. Even if therapy begins immediately after diagnosis, some cancer cells may persist in the body, increasing the risk of a recurrence. Metastasis and recurrence are the leading causes of death from breast cancer. Therefore, detecting a return of breast cancer early has become a pressing medical issue. Evaluating and contrasting various Machine Learning (ML) techniques for breast cancer and recurrence prediction is crucial to choosing the best successful method. Inaccurate forecasts are common when using datasets with a large number of attributes. This study addresses the need for effective feature selection and optimization methods by introducing Recursive Feature Elimination (RFE) and Grey Wolf Optimizer (GWO), in response to the limitations observed in existing approaches. In this research, the performance evaluation of methods is enhanced by employing the RFE and GWO, considering the Wisconsin Diagnostic Breast Cancer (WDBC) and Wisconsin Prognostic Breast Cancer (WPBC) datasets taken from the UCI-ML repository. Various preprocessing techniques are applied to raw data, including imputation, scaling, and others. In the second step, relevant feature correlations are used with RFE to narrow down candidate discriminative features. The GWO chooses the best possible combination of attributes for the most accurate result in the next step. We use seven ML classifiers in both datasets to make a binary decision. On the WDBC and WPBC datasets, several experiments have shown accuracies of 98.25% and 93.27%, precisions of 98.13% and 95.56%, sensitivities of 99.06% and 96.63%, specificities of 96.92% and 73.33%, F1-scores of 98.59% and 96.09% and AUCs of 0.982 and 0.936, respectively. The hybrid approach's superior feature selection improved the accuracy of breast cancer performance indicators and recurrence classification.

MENet: A Mitscherlich function based ensemble of CNN models to classify lung cancer using CT scans

Lung cancer is one of the leading causes of cancer-related deaths worldwide. To reduce the mortality rate, early detection and proper treatment should be ensured. Computer-aided diagnosis methods analyze different modalities of medical images to increase diagnostic precision. In this paper, we propose an ensemble model, called the Mitscherlich function-based Ensemble Network (MENet), which combines the prediction probabilities obtained from three deep learning models, namely Xception, InceptionResNetV2, and MobileNetV2, to improve the accuracy of a lung cancer prediction model. The ensemble approach is based on the Mitscherlich function, which produces a fuzzy rank to combine the outputs of the said base classifiers. The proposed method is trained and tested on the two publicly available lung cancer datasets, namely Iraq-Oncology Teaching Hospital/National Center for Cancer Diseases (IQ-OTH/NCCD) and LIDC-IDRI, both of these are computed tomography (CT) scan datasets. The obtained results in terms of some standard metrics show that the proposed method performs better than state-of-the-art methods. The codes for the proposed work are available at https://github.com/SuryaMajumder/MENet.

Factors influencing clinician and patient interaction with machine learning-based risk prediction models: a systematic review

Machine learning (ML)-based risk prediction models hold the potential to support the health-care setting in several ways; however, use of such models is scarce. We aimed to review health-care professional (HCP) and patient perceptions of ML risk prediction models in published literature, to inform future risk prediction model development. Following database and citation searches, we identified 41 articles suitable for inclusion. Article quality varied with qualitative studies performing strongest. Overall, perceptions of ML risk prediction models were positive. HCPs and patients considered that models have the potential to add benefit in the health-care setting. However, reservations remain; for example, concerns regarding data quality for model development and fears of unintended consequences following ML model use. We identified that public views regarding these models might be more negative than HCPs and that concerns (eg, extra demands on workload) were not always borne out in practice. Conclusions are tempered by the low number of patient and public studies, the absence of participant ethnic diversity, and variation in article quality. We identified gaps in knowledge (particularly views from under-represented groups) and optimum methods for model explanation and alerts, which require future research.

A weighted distance-based dynamic ensemble regression framework for gastric cancer survival time prediction

Accurate prediction of gastric cancer patient survival time is essential for clinical decision-making. However, unified static models lack specificity and flexibility in predictions owing to the varying survival outcomes among gastric cancer patients. We address these problems by using an ensemble learning approach and adaptively assigning greater weights to similar patients to make more targeted predictions when predicting an individual's survival time. We treat these problems as regression problems and introduce a weighted dynamic ensemble regression framework. To better identify similar patients, we devise a method to measure patient similarity, considering the diverse impacts of features. Subsequently, we use this measure to design both a weighted K-means clustering method and a fuzzy K-means sampling technique to group patients and train corresponding base regressors. To achieve more targeted predictions, we calculate the weight of each base regressor based on the similarity between the patient to be predicted and the patient clusters, culminating in the integration of the results. The model is validated on a dataset of 7791 patients, outperforming other models in terms of three evaluation metrics, namely, the root mean square error, mean absolute error, and the coefficient of determination. The weighted dynamic ensemble regression strategy can improve the baseline model by 1.75%, 2.12%, and 13.45% in terms of the three respective metrics while also mitigating the imbalanced survival time distribution issue. This enhanced performance has been statistically validated, even when tested on six public datasets with different sizes. By considering feature variations, patients with distinct survival profiles can be effectively differentiated, and the model predictive performance can be enhanced. The results generated by our proposed model can be invaluable in guiding decisions related to treatment plans and resource allocation. Furthermore, the model has the potential for broader applications in prognosis for other types of cancers or similar regression problems in various domains.

A human-interpretable machine learning pipeline based on ultrasound to support leiomyosarcoma diagnosis

The preoperative evaluation of myometrial tumors is essential to avoid delayed treatment and to establish the appropriate surgical approach. Specifically, the differential diagnosis of leiomyosarcoma (LMS) is particularly challenging due to the overlapping of clinical, laboratory and ultrasound features between fibroids and LMS. In this work, we present a human-interpretable machine learning (ML) pipeline to support the preoperative differential diagnosis of LMS from leiomyomas, based on both clinical data and gynecological ultrasound assessment of 68 patients (8 with LMS diagnosis). The pipeline provides the following novel contributions: (i) end-users have been involved both in the definition of the ML tasks and in the evaluation of the overall approach; (ii) clinical specialists get a full understanding of both the decision-making mechanisms of the ML algorithms and the impact of the features on each automatic decision. Moreover, the proposed pipeline addresses some of the problems concerning both the imbalance of the two classes by analyzing and selecting the best combination of the synthetic oversampling strategy of the minority class and the classification algorithm among different choices, and the explainability of the features at global and local levels. The results show very high performance of the best strategy (AUC = 0.99, F1 = 0.87) and the strong and stable impact of two ultrasound-based features (i.e., tumor borders and consistency of the lesions). Furthermore, the SHAP algorithm was exploited to quantify the impact of the features at the local level and a specific module was developed to provide a template-based natural language (NL) translation of the explanations for enhancing their interpretability and fostering the use of ML in the clinical setting.

Predictive value of machine learning for breast cancer recurrence: a systematic review and meta-analysis

PURPOSE

Recurrence of breast cancer leads to a high lifetime risk and a low 5 year survival rate. Researchers have used machine learning to predict the risk of recurrence in patients with breast cancer, but the predictive performance of machine learning remains controversial. Hence, this study aimed to explore the accuracy of machine learning in predicting breast cancer recurrence risk and aggregate predictive variables to provide guidance for the development of subsequent risk scoring systems.

METHODS

We searched Pubmed, EMBASE, Cochrane, and Web of Science. The risk of bias in the included studies was evaluated using prediction model risk of bias assessment tool (PROBAST). Meta-regression was adopted to explore whether there was a significant difference in the recurrence time by machine learning.

RESULTS

Thirty-four studies involving 67,560 subjects were included, among whom 8695 experienced breast cancer recurrence. The c-index of prediction models was 0.814 (95%CI 0.802-0.826) and 0.770 (95%CI 0.737-0.803) in the training and validation sets, respectively; the sensitivity and specificity were 0.69 (95% CI 0.64-0.74), 0.89 (95% CI 0.86-0.92) in the training, and 0.64 (95% CI 0.58-0.70), 0.88 (95% CI 0.82-0.92) in the validation, respectively. Age, histological grading, and lymph node status are the most commonly used variables in model construction. Attention should be paid to unhealthy lifestyles such as drinking, smoking and BMI as modeling variables. Risk prediction models based on machine learning have long-term monitoring value for breast cancer population, and subsequent studies should consider using large-sample and multi-center data to establish risk equations for verification.

CONCLUSION

Machine learning may be used as a predictive tool for breast cancer recurrence. Currently, there is a lack of effective and universally applicable machine learning models in clinical practice. We expect to incorporate multi-center studies in the future and attempt to develop tools for predicting breast cancer recurrence risk, so as to effectively identify populations at high risk of recurrence and develop personalized follow-up strategies and prognostic interventions to reduce the risk of recurrence.

Evolution of Breast Cancer Recurrence Risk Prediction: A Systematic Review of Statistical and Machine Learning-Based Models

PURPOSE

Selection of appropriate adjuvant therapy to ultimately reduce the risk of breast cancer (BC) recurrence is a challenge for medical oncologists. Several automated risk prediction models have been developed using retrospective clinical data and have evolved significantly over the years in terms of predictors of recurrence, data usage, and predictive techniques (statistical/machine learning [ML]).

METHODS

Following PRISMA guidelines, we performed a systematic literature review of the aforementioned statistical and ML models published between January 2008 and December 2022 through searching five digital databases-PubMed, ScienceDirect, Scopus, Cochrane, and Web of Science. The comprehensive search yielded a total of 163 papers and after a screening process focusing on papers that dealt exclusively with statistical/ML methods, only 23 papers were deemed appropriate for further analysis. We benchmarked the studies on the basis of development, evaluation metrics, and validation strategy with an added emphasis on racial diversity of patients included in the studies.

RESULTS

In total, 30.4% of the included studies use statistical techniques, while 69.6% are ML-based. Among these, traditional ML models (support vector machines, decision tree, logistic regression, and naïve Bayes) are the most frequently used (26.1%) along with deep learning (26.1%). Deep learning and ensemble learning provide the most accurate predictions (AUC = 0.94 each).

CONCLUSION

ML-based prediction models exhibit outstanding performance, yet their practical applicability might be hindered by limited interpretability and reduced generalization. Moreover, predictive models for BC recurrence often focus on limited variables related to tumor, treatment, molecular, and clinical features. Imbalanced classes and the lack of open-source data sets impede model development and validation. Furthermore, existing models predominantly overlook African and Middle Eastern populations, as they are trained and validated mainly on Caucasian and Asian patients.

Internet-Based Healthcare Knowledge Service for Improvement of Chinese Medicine Healthcare Service Quality

With the development of new-generation information technology and increasing health needs, the requirements for Chinese medicine (CM) services have shifted toward the 5P medical mode, which emphasizes preventive, predictive, personalized, participatory, and precision medicine. This implies that CM knowledge services need to be smarter and more sophisticated. This study adopted a bibliometric approach to investigate the current state of development of CM knowledge services, and points out that accurate knowledge service is an inevitable requirement for the modernization of CM. We summarized the concept of smart CM knowledge services and highlighted its main features, including medical homogeneity, knowledge service intelligence, integration of education and research, and precision medicine. Additionally, we explored the intelligent service method of traditional Chinese medicine under the 5P medical mode to support CM automatic knowledge organization and safe sharing, human-machine collaborative knowledge discovery and personalized dynamic knowledge recommendation. Finally, we summarized the innovative modes of CM knowledge services. Our research will guide the quality assurance and innovative development of the traditional Chinese medicine knowledge service model in the era of digital intelligence.

Predicting second breast cancer among women with primary breast cancer using machine learning algorithms, a population-based observational study

Breast cancer survivors often experience recurrence or a second primary cancer. We developed an automated approach to predict the occurrence of any second breast cancer (SBC) using patient-level data and explored the generalizability of the models with an external validation data source. Breast cancer patients from the cancer registry of Zurich, Zug, Schaffhausen, Schwyz (N = 3213; training dataset) and the cancer registry of Ticino (N = 1073; external validation dataset), diagnosed between 2010 and 2018, were used for model training and validation, respectively. Machine learning (ML) methods, namely a feed-forward neural network (ANN), logistic regression, and extreme gradient boosting (XGB) were employed for classification. The best-performing model was selected based on the receiver operating characteristic (ROC) curve. Key characteristics contributing to a high SBC risk were identified. SBC was diagnosed in 6% of all cases. The most important features for SBC prediction were age at incidence, year of birth, stage, and extent of the pathological primary tumor. The ANN model had the highest area under the ROC curve with 0.78 (95% confidence interval [CI] 0.750.82) in the training data and 0.70 (95% CI 0.61-0.79) in the external validation data. Investigating the generalizability of different ML algorithms, we found that the ANN generalized better than the other models on the external validation data. This research is a first step towards the development of an automated tool that could assist clinicians in the identification of women at high risk of developing an SBC and potentially preventing it.

CED: A case-level explainable paramedical diagnosis via AdaGBDT

OBJECTIVE

The rapid growth of medical data has greatly promoted the wide exploitation of machine learning for paramedical diagnosis. Inversely proportional to their performance, most machine learning models generally suffer from the lack of explainability, especially the local explainability of the model, that is, the case-specific explainability.

MATERIALS AND METHODS

In this paper, we proposed a GBDT (Gradient Boosting Decision Tree)-based explainable model for case-specific paramedical diagnostics, and mainly make the following contributions: (1) an adaptive gradient boosting decision tree (AdaGBDT) model is proposed to boost the path-mining for decision effectively; (2) to learn a case-specific feature importance embedding for a specific patient, the bi-side mutual information is applied to characterize the backtracking on the decision path; (3) through the collaborative decision-making by globally explainable AdaGBDT with case-based reasoning (CBR) in the case-specific metric space, some hard cases can be identified by the means of visualized interpretation. The performance of our model is evaluated on the Wisconsin diagnostic breast cancer dataset and the UCI heart disease dataset.

RESULTS

Experiments conducted on two datasets show that our AdaGBDT achieves the best performance, with the F1-value of 0.9647 and 0.8405 respectively. Moreover, a series of experimental analyses and case studies further illustrate the excellent performance of feature importance embedding.

CONCLUSION

The proposed case-specific explainable paramedical diagnosis via AdaGBDT has excellent predictive performance, with both promising case-level and consistent global explainability.

Explainable diabetes classification using hybrid Bayesian-optimized TabNet architecture

Diabetes is a deadly chronic disease that occurs when the pancreas is not able to produce ample insulin or when the body cannot use insulin effectively. If undetected, it may lead to a host of health complications. Hence, accurate and explainable early-stage detection of diabetes is essential for the proper administration of treatment options in leading a healthy and productive life. For this, we developed an interpretable TabNet model tuned via Bayesian optimization (BO). To achieve model-specific interpretability, the attention mechanism of TabNet architecture was used, which offered the local and global model explanations on the influence of the attributes on the outcomes. The model was further explained locally and globally using more robust model-agnostic LIME and SHAP eXplainable Artificial Intelligence (XAI) tools. The proposed model outperformed all benchmarked models by obtaining high accuracy of 92.2% and 99.4% using the Pima Indians diabetes dataset (PIDD) and the early-stage diabetes risk prediction dataset (ESDRPD), respectively. Based on the XAI results, it was clear that the most influential attribute for diabetes classification using PIDD and ESDRPD were Insulin and Polyuria, respectively. The feature importance values registered for insulin was 0.301 (PIDD) and for polyuria 0.206 was registered (ESDRPD). The high accuracy and ancillary interpretability of our objective model is expected to increase end-users trust and confidence in early-stage detection of diabetes.

Explaining predictive factors in patient pathways using autoencoders

This paper introduces an end-to-end methodology to predict a pathway-related outcome and identifying predictive factors using autoencoders. A formal description of autoencoders for explainable binary predictions is presented, along with two objective functions that allows for filtering and inverting negative examples during training. A methodology to model and transform complex medical event logs is also proposed, which keeps the pathway information in terms of events and time, as well as the hierarchy information carried in medical codes. A case study is presented, in which the short-term mortality after the implementation of an Implantable Cardioverter-Defibrillator is predicted. Proposed methodologies have been tested and compared to other predictive methods, both explainable and not explainable. Results show the competitiveness of the method in terms of performances, particularly the use of a Variational Auto Encoder with an inverse objective function. Finally, the explainability of the method has been demonstrated, allowing for the identification of interesting predictive factors validated using relative risks.

A manifesto on explainability for artificial intelligence in medicine

The rapid increase of interest in, and use of, artificial intelligence (AI) in computer applications has raised a parallel concern about its ability (or lack thereof) to provide understandable, or explainable, output to users. This concern is especially legitimate in biomedical contexts, where patient safety is of paramount importance. This position paper brings together seven researchers working in the field with different roles and perspectives, to explore in depth the concept of explainable AI, or XAI, offering a functional definition and conceptual framework or model that can be used when considering XAI. This is followed by a series of desiderata for attaining explainability in AI, each of which touches upon a key domain in biomedicine.

Present and future of machine learning in breast surgery: systematic review

BACKGROUND

Machine learning is a set of models and methods that can automatically detect patterns in vast amounts of data, extract information, and use it to perform decision-making under uncertain conditions. The potential of machine learning is significant, and breast surgeons must strive to be informed with up-to-date knowledge and its applications.

METHODS

A systematic database search of Embase, MEDLINE, the Cochrane database, and Google Scholar, from inception to December 2021, was conducted of original articles that explored the use of machine learning and/or artificial intelligence in breast surgery in EMBASE, MEDLINE, Cochrane database and Google Scholar.

RESULTS

The search yielded 477 articles, of which 14 studies were included in this review, featuring 73 847 patients. Four main areas of machine learning application were identified: predictive modelling of surgical outcomes; breast imaging-based context; screening and triaging of patients with breast cancer; and as network utility for detection. There is evident value of machine learning in preoperative planning and in providing information for surgery both in a cancer and an aesthetic context. Machine learning outperformed traditional statistical modelling in all studies for predicting mortality, morbidity, and quality of life outcomes. Machine learning patterns and associations could support planning, anatomical visualization, and surgical navigation.

CONCLUSION

Machine learning demonstrated promising applications for improving breast surgery outcomes and patient-centred care. Neveretheless, there remain important limitations and ethical concerns relating to implementing artificial intelligence into everyday surgical practices.

Profiling hearing aid users through big data explainable artificial intelligence techniques

Debilitating hearing loss (HL) affects ~6% of the human population. Only 20% of the people in need of a hearing assistive device will eventually seek and acquire one. The number of people that are satisfied with their Hearing Aids (HAids) and continue using them in the long term is even lower. Understanding the personal, behavioral, environmental, or other factors that correlate with the optimal HAid fitting and with users' experience of HAids is a significant step in improving patient satisfaction and quality of life, while reducing societal and financial burden. In SMART BEAR we are addressing this need by making use of the capacity of modern HAids to provide dynamic logging of their operation and by combining this information with a big amount of information about the medical, environmental, and social context of each HAid user. We are studying hearing rehabilitation through a 12-month continuous monitoring of HL patients, collecting data, such as participants' demographics, audiometric and medical data, their cognitive and mental status, their habits, and preferences, through a set of medical devices and wearables, as well as through face-to-face and remote clinical assessments and fitting/fine-tuning sessions. Descriptive, AI-based analysis and assessment of the relationships between heterogeneous data and HL-related parameters will help clinical researchers to better understand the overall health profiles of HL patients, and to identify patterns or relations that may be proven essential for future clinical trials. In addition, the future state and behavioral (e.g., HAids Satisfiability and HAids usage) of the patients will be predicted with time-dependent machine learning models to assist the clinical researchers to decide on the nature of the interventions. Explainable Artificial Intelligence (XAI) techniques will be leveraged to better understand the factors that play a significant role in the success of a hearing rehabilitation program, constructing patient profiles. This paper is a conceptual one aiming to describe the upcoming data collection process and proposed framework for providing a comprehensive profile for patients with HL in the context of EU-funded SMART BEAR project. Such patient profiles can be invaluable in HL treatment as they can help to identify the characteristics making patients more prone to drop out and stop using their HAids, using their HAids sufficiently long during the day, and being more satisfied by their HAids experience. They can also help decrease the number of needed remote sessions with their Audiologist for counseling, and/or HAids fine tuning, or the number of manual changes of HAids program (as indication of poor sound quality and bad adaptation of HAids configuration to patients' real needs and daily challenges), leading to reduced healthcare cost.

Artificial intelligence empowered Digital Health Technologies in Cancer Survivorship Care: a scoping review

Objective

The objectives of this systematic review are to describe features and specific application scenarios for current cancer survivorship care services of Artificial intelligence (AI)-driven digital health technologies (DHTs) and to explore the acceptance and briefly evaluate its feasibility in the application process.

Methods

Search for literatures published from 2010 to 2022 on sites MEDLINE, IEEE-Xplor, PubMed, Embase, Cochrane Central Register of Controlled Trials and Scopus systematically. The types of literatures include original research, descriptive study, randomized controlled trial, pilot study, and feasible or acceptable study. The literatures above described current status and effectiveness of digital medical technologies based on AI and used in cancer survivorship care services. Additionally, we use QuADS quality assessment tool to evaluate the quality of literatures included in this review.

Results

43 studies that met the inclusion criteria were analyzed and qualitatively synthesized. The current status and results related to the application of AI-driven DHTs in cancer survivorship care were reviewed. Most of these studies were designed specifically for breast cancer survivors’ care and focused on the areas of recurrence or secondary cancer prediction, clinical decision support, cancer survivability prediction, population or treatment stratified, anti-cancer treatment-induced adverse reaction prediction, and so on. Applying AI-based DHTs to cancer survivors actually has shown some positive outcomes, including increased motivation of patient-reported outcomes (PROs), reduce fatigue and pain levels, improved quality of life, and physical function. However, current research mostly explored the technology development and formation (testing) phases, with limited-scale population, and single-center trial. Therefore, it is not suitable to draw conclusions that the effectiveness of AI-based DHTs in supportive cancer care, as most of applications are still in the early stage of development and feasibility testing.

Conclusions

While digital therapies are promising in the care of cancer patients, more high-quality studies are still needed in the future to demonstrate the effectiveness of digital therapies in cancer care. Studies should explore how to develop uniform standards for measuring patient-related outcomes, ensure the scientific validity of research methods, and emphasize patient and health practitioner involvement in the development and use of technology.

A two-stage stacked-based heterogeneous ensemble learning for cancer survival prediction

Cancer survival prediction is one of the three major tasks of cancer prognosis. To improve the accuracy of cancer survival prediction, in this paper, we propose a priori knowledge- and stability-based feature selection (PKSFS) method and develop a novel two-stage heterogeneous stacked ensemble learning model (BQAXR) to predict the survival status of cancer patients. Specifically, PKSFS first obtains the optimal feature subsets from the high-dimensional cancer datasets to guide the subsequent model construction. Then, BQAXR seeks to generate five high-quality heterogeneous learners, among which the shortcomings of the learners are overcome by using improved methods, and integrate them in two stages through the stacked generalization strategy based on optimal feature subsets. To verify the merits of PKSFS and BQAXR, this paper collected the real survival datasets of gastric cancer and skin cancer from the Surveillance, Epidemiology, and End Results (SEER) database of the National Cancer Institute, and conducted extensive numerical experiments from different perspectives based on these two datasets. The accuracy and AUC of the proposed method are 0.8209 and 0.8203 in the gastric cancer dataset, and 0.8336 and 0.8214 in the skin cancer dataset. The results show that PKSFS has marked advantages over popular feature selection methods in processing high-dimensional datasets. By taking full advantage of heterogeneous high-quality learners, BQAXR is not only superior to mainstream machine learning methods, but also outperforms improved machine learning methods, which indicates can effectively improve the accuracy of cancer survival prediction and provide a reference for doctors to make medical decisions.

A survey on the interpretability of deep learning in medical diagnosis

Deep learning has demonstrated remarkable performance in the medical domain, with accuracy that rivals or even exceeds that of human experts. However, it has a significant problem that these models are "black-box" structures, which means they are opaque, non-intuitive, and difficult for people to understand. This creates a barrier to the application of deep learning models in clinical practice due to lack of interpretability, trust, and transparency. To overcome this problem, several studies on interpretability have been proposed. Therefore, in this paper, we comprehensively review the interpretability of deep learning in medical diagnosis based on the current literature, including some common interpretability methods used in the medical domain, various applications with interpretability for disease diagnosis, prevalent evaluation metrics, and several disease datasets. In addition, the challenges of interpretability and future research directions are also discussed here. To the best of our knowledge, this is the first time that various applications of interpretability methods for disease diagnosis have been summarized.

Trust-Based Research: Influencing Factors of Patients’ Medical Choice Behavior in the Online Medical Community

The medical service is a special credit commodity, and trust plays a very important role in patients’ online medical choice behavior. By collecting information about the doctors on China’s leading online medical platform (Platform A), a regression analysis model was constructed, based on the credibility theory model, which has the following three dimensions: ability trust, benevolence trust, and integrity trust. The results showed that the medical title of the doctors, their department’s reputation, the number of gifts given to them, and the number of patients who registered with them after diagnosis, among other factors, had a significant, positive impact on the behavior of choosing doctors. Among these considerations, the number of patients registered after diagnosis had the greatest impact on the behavior of choosing doctors. This factor is the result of each doctor’s personal brand management, which reflects their comprehensive ability, reputation and integrity. Compared with previous studies, this paper creatively analyzed the important influence of departmental reputation and the number of patients registered after diagnosis on medical choice behavior and puts forward that a doctor can use the number of patients registered after diagnosis to manage their personal brand. Based on the results of this study, we will also put forward suggestions from the perspectives of patients, doctors and the online medical community.

Efficacy prediction of noninvasive ventilation failure based on the stacking ensemble algorithm and autoencoder

Background

Early prediction of noninvasive ventilation failure is of great significance for critically ill ICU patients to escalate or change treatment. Because clinically collected data are highly time-series correlated and have imbalanced classes, it is difficult to accurately predict the efficacy of noninvasive ventilation for severe patients. This paper aims to precisely predict the failure probability of noninvasive ventilation before or in the early stage (1–2 h) of using it on patients and to explain the correlation of the predicted results.

Methods

In this paper, we proposed a SMSN model (stacking and modified SMOTE algorithm of prediction of noninvasive ventilation failure). In the feature generation stage, we used an autoencoder algorithm based on long short-term memory (LSTM) to automatically extract time series features. In the modelling stage, we adopted a modified SMOTE algorithm to address imbalanced classes, and three classifiers (logistic regression, random forests, and Catboost) were combined with the stacking ensemble algorithm to achieve high prediction accuracy.

Results

Data from 2495 patients were used to train the SMSN model. Among them, 80% of 2495 patients (1996 patients) were randomly selected as the training set, and 20% of these patients (499 patients) were chosen as the testing set. The F1 of the proposed SMSN model was 79.4%, and the accuracy was 88.2%. Compared with the traditional logistic regression algorithm, the F1 and accuracy were improved by 4.7% and 1.3%, respectively.

Conclusions

Through SHAP analysis, oxygenation index, pH and H1FIO₂ collected after 1 h of noninvasive ventilation were the most relevant features affecting the prediction.

A Smart Healthcare Knowledge Service Framework for Hierarchical Medical Treatment System

This paper reveals the research hotspots and development directions of case-based reasoning in the field of health care, and proposes the framework and key technologies of medical knowledge service systems based on case-based reasoning (CBR) in the big data environment. The 2124 articles on medical CBR in the Web of Science were visualized and analyzed using a bibliometrics method, and a CBR-based knowledge service system framework was constructed in the medical Internet of all people, things and data resources environment. An intelligent construction method for the clinical medical case base and the gray case knowledge reasoning model were proposed. A cloud-edge collaboration knowledge service system was developed and applied in a pilot project. Compared with other diagnostic tools, the system provides case-based explanations for its predicted results, making it easier for physicians to understand and accept, so that they can make better decisions. The results show that the system has good interpretability, has better acceptance than the common intelligent decision support system, and strongly supports physician auxiliary diagnosis and treatment as well as clinical teaching.

An Adaptive Deep Ensemble Learning Method for Dynamic Evolving Diagnostic Task Scenarios

Increasingly, machine learning methods have been applied to aid in diagnosis with good results. However, some complex models can confuse physicians because they are difficult to understand, while data differences across diagnostic tasks and institutions can cause model performance fluctuations. To address this challenge, we combined the Deep Ensemble Model (DEM) and tree-structured Parzen Estimator (TPE) and proposed an adaptive deep ensemble learning method (TPE-DEM) for dynamic evolving diagnostic task scenarios. Different from previous research that focuses on achieving better performance with a fixed structure model, our proposed model uses TPE to efficiently aggregate simple models more easily understood by physicians and require less training data. In addition, our proposed model can choose the optimal number of layers for the model and the type and number of basic learners to achieve the best performance in different diagnostic task scenarios based on the data distribution and characteristics of the current diagnostic task. We tested our model on one dataset constructed with a partner hospital and five UCI public datasets with different characteristics and volumes based on various diagnostic tasks. Our performance evaluation results show that our proposed model outperforms other baseline models on different datasets. Our study provides a novel approach for simple and understandable machine learning models in tasks with variable datasets and feature sets, and the findings have important implications for the application of machine learning models in computer-aided diagnosis.

A Personalized Medical Decision Support System Based on Explainable Machine Learning Algorithms and ECC Features: Data from the Real World

Artificial intelligence can help physicians improve the accuracy of breast cancer diagnosis. However, the effectiveness of AI applications is limited by doctors’ adoption of the results recommended by the personalized medical decision support system. Our primary purpose is to study the impact of external case characteristics (ECC) on the effectiveness of the personalized medical decision support system for breast cancer assisted diagnosis (PMDSS-BCAD) in making accurate recommendations. Therefore, we designed a novel comprehensive framework for case-based reasoning (CBR) that takes the impact of external features of cases into account, made use of the naive Bayes and k-nearest neighbor (KNN) algorithms (CBR-ECC), and developed a PMDSS-BCAD system by using the CBR-ECC model and external features as system components. Under the new case-based reasoning framework, the accuracy of the combined model of naive Bayes and KNN with an optimal K value of 2 is 99.40%. Moreover, in a real hospital scenario, users rated the PMDSS-BCAD system, which takes into account the external characteristics of the case, better than the original personalized system. These results suggest that PMDSS-BCD can not only provide doctors with more personalized and accurate results for auxiliary diagnosis, but also improve doctors’ trust in the results, so as to encourage doctors to adopt the results recommended by the personalized system.

Explainable Artificial Intelligence Reveals Novel Insight into Tumor Microenvironment Conditions Linked with Better Prognosis in Patients with Breast Cancer

We investigated the data-driven relationship between immune cell composition in the tumor microenvironment (TME) and the ≥5-year survival rates of breast cancer patients using explainable artificial intelligence (XAI) models. We acquired TCGA breast invasive carcinoma data from the cbioPortal and retrieved immune cell composition estimates from bulk RNA sequencing data from TIMER2.0 based on EPIC, CIBERSORT, TIMER, and xCell computational methods. Novel insights derived from our XAI model showed that B cells, CD8+ T cells, M0 macrophages, and NK T cells are the most critical TME features for enhanced prognosis of breast cancer patients. Our XAI model also revealed the inflection points of these critical TME features, above or below which ≥5-year survival rates improve. Subsequently, we ascertained the conditional probabilities of ≥5-year survival under specific conditions inferred from the inflection points. In particular, the XAI models revealed that the B cell fraction (relative to all cells in a sample) exceeding 0.025, M0 macrophage fraction (relative to the total immune cell content) below 0.05, and NK T cell and CD8+ T cell fractions (based on cancer type-specific arbitrary units) above 0.075 and 0.25, respectively, in the TME could enhance the ≥5-year survival in breast cancer patients. The findings could lead to accurate clinical predictions and enhanced immunotherapies, and to the design of innovative strategies to reprogram the breast TME.

Convolutional Neural Networks for the evaluation of cancer in Barrett's esophagus: Explainable AI to lighten up the black-box

Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their level of accountability and transparency must be provided in such evaluations. The reliability related to machine learning predictions must be explained and interpreted, especially if diagnosis support is addressed. For this task, the black-box nature of deep learning techniques must be lightened up to transfer its promising results into clinical practice. Hence, we aim to investigate the use of explainable artificial intelligence techniques to quantitatively highlight discriminative regions during the classification of early-cancerous tissues in Barrett's esophagus-diagnosed patients. Four Convolutional Neural Network models (AlexNet, SqueezeNet, ResNet50, and VGG16) were analyzed using five different interpretation techniques (saliency, guided backpropagation, integrated gradients, input × gradients, and DeepLIFT) to compare their agreement with experts' previous annotations of cancerous tissue. We could show that saliency attributes match best with the manual experts' delineations. Moreover, there is moderate to high correlation between the sensitivity of a model and the human-and-computer agreement. The results also lightened that the higher the model's sensitivity, the stronger the correlation of human and computational segmentation agreement. We observed a relevant relation between computational learning and experts' insights, demonstrating how human knowledge may influence the correct computational learning.

A new approach for interpretability and reliability in clinical risk prediction: Acute coronary syndrome scenario

INTRODUCTION

The risk prediction of the occurrence of a clinical event is often based on conventional statistical procedures, through the implementation of risk score models. Recently, approaches based on more complex machine learning (ML) methods have been developed. Despite the latter usually have a better predictive performance, they obtain little approval from the physicians, as they lack interpretability and, therefore, clinical confidence. One clinical issue where both types of models have received great attention is the mortality risk prediction after acute coronary syndromes (ACS).

OBJECTIVE

We intend to create a new risk assessment methodology that combines the best characteristics of both risk score and ML models. More specifically, we aim to develop a method that, besides having a good performance, offers a personalized model and outcome for each patient, presents high interpretability, and incorporates an estimation of the prediction reliability which is not usually available. By combining these features in the same approach we expect that it can boost the confidence of physicians to use such a tool in their daily activity.

METHODS

In order to achieve the mentioned goals, a three-step methodology was developed: several rules were created by dichotomizing risk factors; such rules were trained with a machine learning classifier to predict the acceptance degree of each rule (the probability that the rule is correct) for each patient; that information was combined and used to compute the risk of mortality and the reliability of such prediction. The methodology was applied to a dataset of 1111 patients admitted with any type of ACS (myocardial infarction and unstable angina) in two Portuguese hospitals, to assess the 30-days all-cause mortality risk, being validated through a Monte-Carlo cross-validation technique. The performance was compared with state-of-the-art approaches: logistic regression (LR), artificial neural network (ANN), and clinical risk score model (namely the Global Registry of Acute Coronary Events - GRACE).

RESULTS

For the scenario being analyzed, the performance of the proposed approach and the comparison models was assessed through discrimination and calibration. The ability to rank the patients was evaluated through the area under the ROC curve (AUC), and the ability to stratify the patients into low or high-risk groups was determined using the geometric mean (GM) of specificity and sensitivity, the negative predictive value (NPV) and the positive predictive value (PPV). The validation calibration curves were also inspected. The proposed approach (AUC = 81%, GM = 74%, PPV = 17%, NPV = 99%) achieved testing results identical to the standard LR model (AUC = 83%, GM = 73%, PPV = 16%, NPV=99%), but offers superior interpretability and personalization; it also significantly outperforms the GRACE risk model (AUC = 79%, GM = 47%, PPV = 13%, NPV = 98%) and the standard ANN model (AUC = 78%, GM = 70%, PPV = 13%, NPV = 98%). The calibration curve also suggests a very good generalization ability of the obtained model as it approaches the ideal curve (slope = 0.96). Finally, the reliability estimation of individual predictions presented a great correlation with the misclassifications rate.

CONCLUSION

We developed and described a new tool that showed great potential to guide the clinical staff in the risk assessment and decision-making process, and to obtain their wide acceptance due to its interpretability and reliability estimation properties. The methodology presented a good performance when applied to ACS events, but those properties may have a beneficial application in other clinical scenarios as well.

A heterogeneous ensemble learning method for neuroblastoma survival prediction

Neuroblastoma is a pediatric cancer with high morbidity and mortality. Accurate survival prediction of patients with neuroblastoma plays an important role in the formulation of treatment plans. In this study, we proposed a heterogeneous ensemble learning method to predict the survival of neuroblastoma patients and extract decision rules from the proposed method to assist doctors in making decisions. After data preprocessing, five heterogeneous base learners were developed, which consisted of decision tree, random forest, support vector machine based on genetic algorithm, extreme gradient boosting and light gradient boosting machine. Subsequently, a heterogeneous feature selection method was devised to obtain the optimal feature subset of each base learner, and the optimal feature subset of each base learner guided the construction of the base learners as a priori knowledge. Furthermore, an area under curve-based ensemble mechanism was proposed to integrate the five heterogeneous base learners. Finally, the proposed method was compared with mainstream machine learning methods from different indicators, and valuable information was extracted by using the partial dependency plot analysis method and rule-extracted method from the proposed method. Experimental results show that the proposed method achieves an accuracy of 91.64%, recall of 91.14%, and AUC of 91.35% and is significantly better than the mainstream machine learning methods. In addition, interpretable rules with accuracy higher than 0.900 and predicted responses are extracted from the proposed method. Our study can effectively improve the performance of the clinical decision support system to improve the survival of neuroblastoma patients.

Similarity measures and attribute selection for case-based reasoning in transcatheter aortic valve implantation

In a clinical decision support system, the purpose of case-based reasoning is to help clinicians make convenient decisions for diagnoses or interventional gestures. Past experience, which is represented by a case-base of previous patients, is exploited to solve similar current problems using four steps-retrieve, reuse, revise, and retain. The proposed case-based reasoning has been focused on transcatheter aortic valve implantation to respond to clinical issues pertaining vascular access and prosthesis choices. The computation of a relevant similarity measure is an essential processing step employed to obtain a set of retrieved cases from a case-base. A hierarchical similarity measure that is based on a clinical decision tree is proposed to better integrate the clinical knowledge, especially in terms of case representation, case selection and attributes weighting. A case-base of 138 patients is used to evaluate the case-based reasoning performance, and retrieve- and reuse-based criteria have been considered. The sensitivity for the vascular access and the prosthesis choice is found to 0.88 and 0.94, respectively, with the use of the hierarchical similarity measure as opposed to 0.53 and 0.79 for the standard similarity measure. Ninety percent of the suggested solutions are correctly classified for the proposed metric when four cases are retrieved. Using a dedicated similarity measure, with relevant and weighted attributes selected through a clinical decision tree, the set of retrieved cases, and consequently, the decision suggested by the case-based reasoning are substantially improved over state-of-the-art similarity measures.