Understanding the bias in machine learning systems for cardiovascular disease risk assessment: The first of its kind review

Suri, Jasjit S.; Bhagawati, Mrinalini; Sauseng, Paul; Protogeron, Athanasios; Sfikakis, Petros P.; Kitas, George D.; Khanna, Narendra N.; Ruzsa, Zoltán; Sharma, Aditya; Saxena, Sanjay; Faa, Gavino; Paraskevas, Kosmas I.; Laird, John R.; Johri, Amer M.; Saba, Luca; Kalra, Manudeep

doi:10.1016/j.compbiomed.2021.105204

Cited by 54 publications

(38 citation statements)

References 100 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…12 One of the drawbacks of ML-based algorithms for assessment of cardiovascular disease risk is any associated bias. 13 Based on good vs moderate vs poor sophisticated use of AI architecture, the supervised vs unsupervised paradigm for developing training ML models, fully vs partially adopting feature extraction methods and complete vs partial vs lack of usage of popular sophisticated models, ML studies can be divided into low-, moderate-and high-bias cluster studies. 13 Several scoring attributes can be used for computing the ML bias and estimating the risk of bias.…”

mentioning

confidence: 99%

“…The overall performance of ML systems can be improved by considering the missing attributes, such as verification and validation. 13 In the last few years, AI has acquired several applications in the diagnosis and treatment of a variety of diseases and conditions. 14,15 As supported by Lareyre et al, 1 AI has extended into several applications in vascular diseases, ranging from diagnosis and classification to prognosis and risk prediction, imaging analysis and treatment.…”

mentioning

confidence: 99%

“…13 Based on good vs moderate vs poor sophisticated use of AI architecture, the supervised vs unsupervised paradigm for developing training ML models, fully vs partially adopting feature extraction methods and complete vs partial vs lack of usage of popular sophisticated models, ML studies can be divided into low-, moderate- and high-bias cluster studies. 13 Several scoring attributes can be used for computing the ML bias and estimating the risk of bias. These can be broadly grouped into 9 clusters: (i) demographics, (ii) AI architecture design, (iii) optimization, (iv) performance parameter, (v) scientific validation, (vi) statistical test, (vii) clinical validation, (viii) clinical evaluation and (ix) survival and hazard analysis.…”

mentioning

confidence: 99%

“…These can be broadly grouped into 9 clusters: (i) demographics, (ii) AI architecture design, (iii) optimization, (iv) performance parameter, (v) scientific validation, (vi) statistical test, (vii) clinical validation, (viii) clinical evaluation and (ix) survival and hazard analysis. 13 Each of these attributes has independent criteria for grading and scoring and can be used to classify the risk of bias. The overall performance of ML systems can be improved by considering the missing attributes, such as verification and validation.…”

mentioning

confidence: 99%

“…The overall performance of ML systems can be improved by considering the missing attributes, such as verification and validation. 13…”

mentioning

confidence: 99%

See 4 more Smart Citations

Applications of Artificial Intelligence in Vascular Diseases

Paraskevas¹,

Saba

Suri³

2022

Angiology

Self Cite

View full text Add to dashboard Cite

In their comprehensive article, Lareyre et al. evaluated scientific publications on artificial intelligence (AI) on noncardiac vascular diseases. 1 The authors provided a quantitative assessment of research output on publications related to AI in carotid artery stenosis, aortic and peripheral artery disease. 1 The bibliometric analysis of original articles published on AI in non-cardiac vascular diseases showed increased numbers over the past 5 years. 1 AI, with its numerous applications, has emerged as an indispensable part of healthcare. By means of its 2 subcategories, namely, machine learning (ML) and deep learning (DL), AI has expanded over the field of medical imaging and medical diagnosis to cover all aspects of vascular disease, from diagnosis to staging and treatment. 2,3 Atherosclerotic plaque formation in the carotid or peripheral arteries is a dynamic process. Advancements in ML and DL have made it possible to identify, classify and characterize the atherosclerotic plaque. 4 The nature of AI design has evolved over time, starting from manual, signal processing methods and eventually leading to AI learning models. 4 ML can be used for plaque characterization based on ultrasound, computed tomography (CT) or magnetic resonance angiography (MRI). While AI models may independently operate on each imaging technique, there are differences and similarities covering various aspects, such as multiresolution, extracted and texture/histogram features, performance metrics, segmentation and classification. 4 Diet and nutrition contribute to cardiovascular events in countries with high socio-demographic index. 5,6 Defining the role of nutrition in the formation of carotid atherosclerosis can determine cerebrovascular/cardiovascular events and all-cause mortality. 5,6 Another way to improve risk assessment is to use AI-based algorithms for cardiovascular risk stratification. 7 Such AI-based algorithms can handle a large number of risk predictors in a model. 7 Furthermore, since AI-based algorithms are data-specific algorithms, they capture and learn from the complex non-linear interrelationship among the risk predictors. 7 Conventional cardiovascular risk calculators (e.g. the Framingham risk score) have certain drawbacks when

show abstract

mentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

“…The overall performance of ML systems can be improved by considering the missing attributes, such as verification and validation. 13…”

mentioning

confidence: 99%

See 3 more Smart Citations

Applications of Artificial Intelligence in Vascular Diseases

Paraskevas¹,

Saba

Suri³

2022

Angiology

Self Cite

View full text Add to dashboard Cite

show abstract

SERS‐Based Biosensors Combined with Machine Learning for Medical Application**

et al. 2023

View full text Add to dashboard Cite

Surface‐enhanced Raman spectroscopy (SERS) has shown strength in non‐invasive, rapid, trace analysis and has been used in many fields in medicine. Machine learning (ML) is an algorithm that can imitate human learning styles and structure existing content with the knowledge to effectively improve learning efficiency. Integrating SERS and ML can have a promising future in the medical field. In this review, we summarize the applications of SERS combined with ML in recent years, such as the recognition of biological molecules, rapid diagnosis of diseases, developing of new immunoassay techniques, and enhancing SERS capabilities in semi‐quantitative measurements. Ultimately, the possible opportunities and challenges of combining SERS with ML are addressed.

show abstract