Developing an Explainable Machine Learning-Based Thyroid Disease Prediction Model

Healthcare and medicine are key areas where machine learning algorithms are widely used. The medical decision support systems thus created are accurate enough, however, they suffer from the lack of transparency in decision making and shows a black box behavior. However, transparency and trust are significant in the field of health and medicine and hence, a black box system is sub optimal in terms of widespread applicability and reach. Hence, the explainablility of the research make the system reliable and understandable, thereby enhancing its social acceptability. The presented work explores a thyroid disease diagnosis system. SHAP, a popular method based on coalition game theory is used for interpretability of results. The work explains the system behavior both locally and globally and shows how machine leaning can be used to ascertain the causality of the disease and support doctors to suggest the most effective treatment of the disease. The work not only demonstrates the results of machine learning algorithms but also explains related feature importance and model insights.

A hybrid approach to medical decision-making: diagnosis of heart disease with machine-learning model

Heart disease is one of the most widely spreading and deadliest diseases across the world. In this study, we have proposed hybrid model for heart disease prediction by employing random forest and support vector machine. With random forest, iterative feature elimination is carried out to select heart disease features that improves predictive outcome of support vector machine for heart disease prediction. Experiment is conducted on the proposed model using test set and the experimental result evidently appears to prove that the performance of the proposed hybrid model is better as compared to an individual random forest and support vector machine. Overall, we have developed more accurate and computationally efficient model for heart disease prediction with accuracy of 98.3%. Moreover, experiment is conducted to analyze the effect of regularization parameter (C) and gamma on the performance of support vector machine. The experimental result evidently reveals that support vector machine is very sensitive to C and gamma.

Relevant Biomarkers in Medical Practices: An Analysis of the Needs Addressed by an International Survey

(1) Backround: Technological advances should foster gains in physicians’ efficiency. For example, a reduction of the medical decision time can be enabled by faster biological tests. The main objective of this study was to collect responses from an international panel of physicians on their needs for biomarkers and also to convey the improvement in the outcome to be made possible by the potential development of fast diagnostic tests for these biomarkers. (2) Methods: we distributed a questionnaire on the Internet to physicians. (3) Results: 508 physicians participated in this survey. The mean age was 38 years. General practice and emergency medicine were heavily represented, with 95% CIs of 44% (39.78, 48.41) and 32% (27.84, 35.94)), respectively. The two most represented countries were France (95% CI: 74% (70.20, 77.83)) and the USA (95% CI: 11% (8.65, 14.18)). Ninety-eight percentages of the physicians thought that obtaining cited biomarkers more quickly would be beneficial to their practice and to patient’s care. The main biomarkers of interest identified by our panel were troponin (95% CI: 51% (46.24, 54.94)), C-reactive protein (95% CI: 42% (38.03, 46.62)), D-dimer (95% CI: 29% (24.80, 32.68)), and brain natriuretic peptide (95% CI: 13% (10.25, 16.13)). (4) Conclusions: Our study highlights the real technological need for fast biomarker results, which could be provided by biosensors. The relevance of some answers such as troponin is questionable.

A Machine Learning Ensemble Based on Radiomics to Predict BI-RADS Category and Reduce the Biopsy Rate of Ultrasound-Detected Suspicious Breast Masses

We developed a machine learning model based on radiomics to predict the BI-RADS category of ultrasound-detected suspicious breast lesions and support medical decision-making towards short-interval follow-up versus tissue sampling. From a retrospective 2015–2019 series of ultrasound-guided core needle biopsies performed by four board-certified breast radiologists using six ultrasound systems from three vendors, we collected 821 images of 834 suspicious breast masses from 819 patients, 404 malignant and 430 benign according to histopathology. A balanced image set of biopsy-proven benign (n = 299) and malignant (n = 299) lesions was used for training and cross-validation of ensembles of machine learning algorithms supervised during learning by histopathological diagnosis as a reference standard. Based on a majority vote (over 80% of the votes to have a valid prediction of benign lesion), an ensemble of support vector machines showed an ability to reduce the biopsy rate of benign lesions by 15% to 18%, always keeping a sensitivity over 94%, when externally tested on 236 images from two image sets: (1) 123 lesions (51 malignant and 72 benign) obtained from two ultrasound systems used for training and from a different one, resulting in a positive predictive value (PPV) of 45.9% (95% confidence interval 36.3–55.7%) versus a radiologists’ PPV of 41.5% (p < 0.005), combined with a 98.0% sensitivity (89.6–99.9%); (2) 113 lesions (54 malignant and 59 benign) obtained from two ultrasound systems from vendors different from those used for training, resulting into a 50.5% PPV (40.4–60.6%) versus a radiologists’ PPV of 47.8% (p < 0.005), combined with a 94.4% sensitivity (84.6–98.8%). Errors in BI-RADS 3 category (i.e., assigned by the model as BI-RADS 4) were 0.8% and 2.7% in the Testing set I and II, respectively. The board-certified breast radiologist accepted the BI-RADS classes assigned by the model in 114 masses (92.7%) and modified the BI-RADS classes of 9 breast masses (7.3%). In six of nine cases, the model performed better than the radiologist did, since it assigned a BI-RADS 3 classification to histopathology-confirmed benign masses that were classified as BI-RADS 4 by the radiologist.

Fermatean Hesitant Fuzzy Sets with Medical Decision Making Application

Fermatean fuzzy set idea obtained by combining fermatean fuzzy sets and hesitant fuzzy sets can be used in practice to simplify the solution of complicated multi-criteria decision-making (MCDM) problems. Initially, the notion of fermatean hesitant fuzzy set is given and the operations related to this concept are presented. Aggregation operators according to fermatean hesitant fuzzy sets are given and basic properties of these operators are studied. To choose the best alternative in practice, a novel MCDM method that is obtained with operators has been created. Finally, an example of infectious diseases was examined to indicate the effectiveness of the suggested techniques.

Affective influences on clinical reasoning and diagnosis: insights from social psychology and new research opportunities

Psychological research consistently demonstrates that affect can play an important role in decision-making across a broad range of contexts. Despite this, the role of affect in clinical reasoning and medical decision-making has received relatively little attention. Integrating the affect, social cognition, and patient safety literatures can provide new insights that promise to advance our understanding of clinical reasoning and lay the foundation for novel interventions to reduce diagnostic errors and improve patient safety. In this paper, we briefly review the ways in which psychologists differentiate various types of affect. We then consider existing research examining the influence of both positive and negative affect on clinical reasoning and diagnosis. Finally, we introduce an empirically supported theoretical framework from social psychology that explains the cognitive processes by which these effects emerge and demonstrates that cognitive interventions can alter these processes. Such interventions, if adapted to a medical context, hold great promise for reducing errors that emerge from faulty thinking when healthcare providers experience different affective responses.

Virtual Medicine, Virtual Disease, Real Consequences

This chapter investigates the hypothesis that the COVID-19 pandemic was the perfect storm due to the misalignment of competing elements of the US healthcare system, the economic commoditization of disease, the economic commoditization of healthcare delivery, and inadequate data to inform medical decision making on a mass scale. The culmination of a decades-long devolution away from patient care to healthcare or more appropriately sick-care created a system that was unable to quickly find the common ground needed to deal with the pandemic known as COVID-19.

Using artificial intelligence in healthcare: Allocating liability and risks

Within the framework of this article, the author considers the features regarding the application and use of artificial intelligence (AI) in medical practice. This includes complex issues related to the personal liability of a doctor when making decisions on diagnostics and treatment based on an algorithm proposal (a system for supporting medical decisions), as well as possible options for the responsibility of the algorithm (AI) developer. The analysis provides an overview of the existing system for holding medical professionals accountable, as well as an assessment of possible options for the distribution of responsibility in connection with the widespread introduction of AI into the work of doctors alongside the possible introduction of AI into standard medical care. The author considers the possibility of establishing more serious requirements for the collection of information on the side effects of such devices for an AI registered as a medical device. Using the method of legal analysis and the comparative legal method, the author analyzes the current global trends in the distribution of responsibility for harm in such cases where there is an error and/or inaccuracy in making a medical decision; as a result of this, the author demonstrates possible options for the distribution of the roles of the healthcare professional and AI in the near future.

A novel assessment method for COVID-19 humoral immunity duration using serial measurements in naturally-infected and vaccinated subjects.

Background: It is crucial for medical decision-making and vaccination strategies to collect information on sustainability of immune responses after infection or vaccination, and how long-lasting antibodies against SARS-COV-2 could provide a humoral and protective immunity, preventing reinfection with SARS-CoV-2 or its variants. The aim of this study is to present a novel method to quantitatively measure and monitor the diversity of SARS-CoV-2 specific antibody profiles over time. Methods: Two collections of serum samples were used in this study: A collection from 20 naturally-infected subjects (follow-ups to 1 year) and a collection from 83 subjects vaccinated with one or two doses of Pfizer BioNtech vaccine (BNT162b2/BNT162b2) (follow-ups to 6 months). The Multi-SARS-CoV-2 assay, a multiparameter serology test, developed for the serological confirmation of past-infections was used to determine the reactivity of six different SARS-CoV-2 antigens. For each patient sample, 3 dilutions (1/50, 1/400 and 1/3200) were defined as an optimal set over the six antigens and their respective linear ranges, allowing accurate quantitation of the corresponding six specific antibodies. Nonlinear mixed-effects modelling was applied to convert intensity readings from 3 determined dilutions to a single quantification value for each antibody. Results: Median half-life for the 20 naturally infected vs 74 vaccinated subjects (two doses) was respectively 120 vs 50 days for RBD, 127 vs 53 days for S1 and 187 vs 86 days for S2 antibodies. Respectively, 90% of the antibody concentration wanes after 158 vs 398 days for RBD, 171 vs 420 days for S1, and 225 vs 620 days for S2 after the second vaccine shot. Conclusion: The newly proposed method, based on a series of a limited number of dilutions, can convert a conventional qualitative assay into a quantitative assay. This conversion helps define the sustainability of specific immune responses against each relevant viral antigen and can help in defining the protection characteristics after an infection or a vaccination.