☆ 4.7 Letter

Machine Learning Compared With Pathologist Assessment

JAMA-JOURNAL OF THE AMERICAN MEDICAL ASSOCIATION (2018)

Journal

JAMA-JOURNAL OF THE AMERICAN MEDICAL ASSOCIATION

Volume 319, Issue 16, Pages 1725-1726

Publisher

AMER MEDICAL ASSOC

DOI: 10.1001/jama.2018.1466

Keywords

-

Categories

Medicine, General & Internal

Ask authors/readers for more resources

Protocol

Community support

Reagent

Community support

Authors

I am an author on this paper

Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7

Not enough ratings

Secondary Ratings

Novelty

-

Significance

-

Scientific rigor

-

Rate this paper

Recommended

Article Medicine, General & Internal

Reliability of CD44, CD24, and ALDH1A1 immunohistochemical staining: Pathologist assessment compared to quantitative image analysis

Lusine Yaghjyan, Yujing J. Heng, Gabrielle M. Baker, Vanessa Bret-Mounet, Divya Murthy, Matt B. Mahoney, Yi Mu, Bernard Rosner, Rulla M. Tamimi

Summary: This study compared the assessment of stem cell markers CD44, CD24, and ALDH1A1 in breast tissue of cancer-free women by pathologists and automated image analysis. The results showed variable correlations between manual and computational assessments, suggesting the use of computational platforms for evaluating stem cell markers in large-scale epidemiological studies.

FRONTIERS IN MEDICINE (2022)

Add to Collection

Article Cardiac & Cardiovascular Systems

Quantitative assessment of myocardial fibrosis by digital image analysis: An adjunctive tool for pathologist ground truth

Joao Abecasis, Nuno Cortez-Dias, Daniel Gomes Pinto, Pedro Lopes, Marcio Madeira, Sancia Ramos, Victor Gil, Nuno Cardim, Ana Felix

Summary: Using an automated artificial intelligence software can improve the quantification of myocardial fibrosis (MF) on endomyocardial biopsies (EMB) and enhance the reliability of diagnostic and prognostic assessments. This study analyzed EMB samples from 30 patients with severe aortic stenosis and compared the results obtained by experienced pathologists with those obtained using an automatic quantification algorithm. The results showed a high correlation between the automatic algorithm and pathologists' quantification, leading to improved inter-observer agreement.

CARDIOVASCULAR PATHOLOGY (2023)

Add to Collection

Article Medicine, Research & Experimental

CT radiomics facilitates more accurate diagnosis of COVID-19 pneumonia: compared with CO-RADS

Huanhuan Liu, Hua Ren, Zengbin Wu, He Xu, Shuhai Zhang, Jinning Li, Liang Hou, Runmin Chi, Hui Zheng, Yanhong Chen, Shaofeng Duan, Huimin Li, Zongyu Xie, Dengbin Wang

Summary: The study demonstrated the superiority of radiomics model in diagnosing COVID-19 pneumonia, facilitating more rapid and accurate detection.

JOURNAL OF TRANSLATIONAL MEDICINE (2021)

Add to Collection

Review Medicine, General & Internal

AI-based digital image dietary assessment methods compared to humans and ground truth: a systematic review

Eleanor Shonkoff, Kelly Copeland Cara, Xuechen (Anna) Pei, Mei Chung, Shreyas Kamath, Karen Panetta, Erin Hennessy

Summary: Overall, the study conducted a systematic review on AI-based automated dietary assessment methods and found that relative errors suggest similar or even better accuracy in volume and calorie estimation compared to human assessment. However, due to variability in food image databases and reported results, meta-analytic synthesis could not be conducted.

ANNALS OF MEDICINE (2023)

Add to Collection

Article Biochemical Research Methods

GAN-based data augmentation for transcriptomics: survey and comparative assessment

Alice Lacan, Michele Sebag, Blaise Hanczar

Summary: This article analyzes the application of GAN-based data augmentation strategies in cancer phenotype classification. The results show a significant improvement in binary and multiclass classification performance through data augmentation. Without augmentation, the accuracy of classifiers trained on only 50 RNA-seq samples is 94% and 70% for binary and tissue classification respectively, while adding 1000 augmented samples increases the accuracy to 98% and 94%. The strength and training cost of the generative models positively correlate with the augmentation performance and generated data quality. Multiple performance indicators are required to assess the quality of the generated data correctly.

BIOINFORMATICS (2023)

Add to Collection

Article Business, Finance

Safeguarding FinTech innovations with machine learning: Comparative assessment of various approaches

Nawazish Mirza, Mohamed Elhoseny, Muhammad Umar, Noura Metawa

Summary: Protecting data is crucial for the development of FinTech, and machine-learning approaches can be used to detect anomalies and predict fraud in FinTech systems. Our assessment of anomaly detection techniques on various databases confirms the effectiveness of machine-learning in fraud detection, although the degree of effectiveness may vary.

RESEARCH IN INTERNATIONAL BUSINESS AND FINANCE (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

Landslide susceptibility assessment using locally weighted learning integrated with machine learning algorithms

Haoyuan Hong

Summary: The aim of this study is to propose five integration models for assessing landslide susceptibility. By integrating different algorithms and models, it is found that the LWL-RS-ADT model is the most reliable and stable. The study also identifies NDVI, lithology, and altitude as key factors in predicting landslide susceptibility.

EXPERT SYSTEMS WITH APPLICATIONS (2024)

Add to Collection

Article Mathematics

The Fama-French Five-Factor Model with Hurst Exponents Compared with Machine Learning Methods

Yicun Li, Yuanyang Teng

Summary: This paper constructs a seven-factor model using monthly data of the A-share market, adding the Hurst exponent factor and the momentum factor to a Fama-French five-factor model. The results show a 7% improvement in the average R-squared. Additionally, five machine learning algorithms are compared with ordinary least squares (OLS) in terms of fitting power, indicating that regularization algorithms have worse performance, while SVM and random forests show good improvement and the efficacy of neural networks varies depending on the data, frequency, and period.

MATHEMATICS (2023)

Add to Collection

Article Radiology, Nuclear Medicine & Medical Imaging

Introducing a gatekeeping system for amyloid status assessment in mild cognitive impairment

E. Doering, M. C. Hoenig, G. N. Bischof, K. P. Bohn, L. M. Ellingsen, T. van Eimeren, A. Drzezga

Summary: This study proposes a machine learning-based gatekeeping system to predict A-status in MCI patients based on pre-existing information on APOE genotype, F-18-FDG PET, age, and sex. The algorithm shows good reliability in estimating A-status and could assist in better assessing individual risk for developing AD and selecting patients for further evaluation. Further potential utility in clinical routine and clinical trials is discussed.

EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING (2022)

Add to Collection

Article Pathology

Multi-institutional Assessment of Pathologist Scoring HER2 Immunohistochemistry

Charles J. Robbins, Aileen I. Fernandez, Gang Han, Serena Wong, Malini Harigopal, Mirna Podoll, Kamaljeet Singh, Amy Ly, M. Gabriela Kuba, Hannah Wen, Mary Ann Sanders, Jane Brock, Shi Wei, Oluwole Fadare, Krisztina Hanley, Julie Jorns, Olivia L. Snir, Esther Yoon, Kim Rabe, T. Rinda Soong, Emily S. Reisenbichler, David L. Rimm

Summary: This study evaluated the concordance and interrater reliability of scoring HER2 IHC in breast cancer biopsies by 18 breast cancer-specialized pathologists. The results showed poor concordance within the intermediate categories and for the IHC 0 cases. The use of a 3-category system partially reduced the discordance. The study also revealed the optimal number of raters for determining HER2 IHC scores of 0 and 3+.

MODERN PATHOLOGY (2023)

Add to Collection

Article Computer Science, Information Systems

Transforming fatigue assessment: Smartphone-based system with digitized motor skill tests

Elli Valla, Ain-Joonas Toose, Sven Nomm, Aaro Toomela

Summary: This paper explores the feasibility of using smartphones to collect motor skill related data for fatigue detection through machine learning. The goal is to gain insights into fatigue and support the development of more effective interventions.

INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS (2023)

Add to Collection

Article Biochemical Research Methods

Do we need different machine learning algorithms for QSAR modeling? A comprehensive assessment of 16 machine learning algorithms on 14 QSAR data sets

Zhenxing Wu, Minfeng Zhu, Yu Kang, Elaine Lai-Han Leung, Tailong Lei, Chao Shen, Dejun Jiang, Zhe Wang, Dongsheng Cao, Tingjun Hou

Summary: A study on learning QSAR models using various ML algorithms for 14 public datasets showed that rbf-SVM, rbf-GPR, XGBoost, and DNN generally perform better than other algorithms. SVM and XGBoost are recommended for regression learning on small datasets, while XGBoost is an excellent choice for large datasets. Ensemble models integrating multiple algorithms can improve prediction accuracy.

BRIEFINGS IN BIOINFORMATICS (2021)

Add to Collection

Article Multidisciplinary Sciences

Phenotype-Based Threat Assessment

Jing Yang, Mohammed Eslami, Yi-Pei Chen, Mayukh Das, Dongmei Zhang, Shaorong Chen, Alexandria-Jade Roberts, Mark Weston, Angelina Volkova, Kasra Faghihi, Robbie K. Moore, Robert C. Alaniz, Alice R. Wattam, Allan Dickerman, Clark Cucinell, Jarred Kendziorski, Sean Coburn, Holly Paterson, Osahon Obanor, Jason Maples, Stephanie Servetas, Jennifer Dootz, Qing-Ming Qin, James E. Samuel, Arum Han, Erin J. van Schaik, Paul de Figueiredo

Summary: This study presents a machine learning approach to assess potential pathogenic threat by analyzing phenotypic hallmarks of bacterial pathogens, offering new strategies for pathogen identification.

PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA (2022)

Add to Collection

Review Biochemical Research Methods

Critical assessment of computational tools for prokaryotic and eukaryotic promoter prediction

Meng Zhang, Cangzhi Jia, Fuyi Li, Chen Li, Yan Zhu, Tatsuya Akutsu, Geoffrey Webb, Quan Zou, Lachlan J. M. Coin, Jiangning Song

Summary: This study provides benchmark datasets for promoter prediction in 58 different species, and finds that deep learning and traditional machine learning-based approaches generally outperform scoring function-based approaches.

BRIEFINGS IN BIOINFORMATICS (2022)

Add to Collection

Article Computer Science, Artificial Intelligence

Credibility assessment of financial stock tweets

Lewis Evans, Majdi Owda, Keeley Crockett, Ana Fernandez Vilas

Summary: Social media, especially Twitter, has been used by investors to discuss stock exchange-listed companies, highlighting the importance of credible information for making well-informed investment decisions. This paper introduces a novel methodology to assess the credibility of financial stock market tweets, demonstrating that combining general and financial features in classifiers can improve performance. Random Forest was found to be the top performer, although it required more features compared to other classifiers.

EXPERT SYSTEMS WITH APPLICATIONS (2021)

Add to Collection

Article Medicine, General & Internal

Assessing Performance and Clinical Usefulness in Prediction Models With Survival Outcomes: Practical Guidance for Cox Proportional Hazards Models

David J. McLernon, Daniele Giardiello, Ben Van Calster, Laure Wynants, Nan van Geloven, Maarten van Smeden, Terry Therneau, Ewout W. Steyerberg, STRATOS Initiative

Summary: Risk prediction models need validation to assess their performance. This article focuses on evaluating predictions and improving clinical decision making using survival models based on Cox proportional hazards regression. The authors present a case study on breast cancer patients, where a Cox regression model is developed and validated for prediction of recurrence or death.

ANNALS OF INTERNAL MEDICINE (2023)

Add to Collection

Article Cardiac & Cardiovascular Systems

Applicability of European Society of Cardiology guidelines according to gross national income

Wouter B. van Dijk, Ewoud Schuit, Rieke van der Graaf, Rolf H. H. Groenwold, Sara Laurijssen, Barbara Casadei, Marco Roffi, Seye Abimbola, Martine C. de Vries, Diederick E. Grobbee

Summary: This study evaluated the feasibility of implementing ESC guidelines on general cardiology in 102 countries and found that compliance varied based on the country's income level. High-income countries had better compliance, while low-income countries had lower compliance, largely due to lack of reimbursement and financial barriers.

EUROPEAN HEART JOURNAL (2023)

Add to Collection

Editorial Material Endocrinology & Metabolism

Is it a risk factor, a predictor, or even both? The multiple faces of multivariable regression analysis

Rolf H. H. Groenwold, Olaf M. Dekkers

Summary: The medical research literature often includes regression analyses with multiple covariates, known as multivariable regression models. However, the interpretation of their results is often unclear or lacks justification. This article outlines different research objectives that justify using multivariable regression analysis, as well as caveats in interpreting the results.

EUROPEAN JOURNAL OF ENDOCRINOLOGY (2023)

Add to Collection

Article Mathematical & Computational Biology

Comparison of likelihood penalization and variance decomposition approaches for clinical prediction models: A simulation study

Anna Lohmann, Rolf H. H. Groenwold, Maarten van Smeden

Summary: Logistic regression is commonly used in developing clinical risk prediction models. Penalization and variance decomposition techniques have been shown to improve predictive performance of the logistic model. In this study, we compared the predictive performance of risk prediction models derived from elastic net, Lasso, ridge, incomplete principal component regression, and incomplete partial least squares regression using simulation. Our results demonstrate that models developed using penalization approaches outperform those developed using maximum likelihood estimation, particularly in terms of calibration.

BIOMETRICAL JOURNAL (2023)

Add to Collection

Article Medicine, General & Internal

There is no such thing as a validated prediction model

Ben Van Calster, Ewout W. W. Steyerberg, Laure Wynants, Maarten van Smeden

Summary: Clinical prediction models should be validated before implementation, as the performance of a model may vary across patient populations and measurement procedures. Validation studies should focus on understanding and quantifying heterogeneity and monitoring performance over time, to ensure that prediction models remain up-to-date and safe for clinical decision-making.

BMC MEDICINE (2023)

Add to Collection

Review Health Care Sciences & Services

Overinterpretation of findings in machine learning prediction model studies in oncology: a systematic review

Paula Dhiman, Jie Ma, Constanza L. Andaur Navarro, Benjamin Speich, Garrett Bullock, Johanna A. A. Damen, Lotty Hooft, Shona Kirtley, Richard D. Riley, Ben Van Calster, Karel G. M. Moons, Gary S. Collins

Summary: This article conducted a systematic review on oncology-related studies that developed and validated prognostic models using machine learning. The findings revealed the presence of spin, i.e., overinterpretation of findings, in these studies. The inconsistent reporting and use of overly strong or leading words in the publications indicate the need for caution when reading and using prognostic models in oncology.

JOURNAL OF CLINICAL EPIDEMIOLOGY (2023)

Add to Collection

Article Cardiac & Cardiovascular Systems

Cardiovascular Risk Prediction in Men and Women Aged Under 50 Years Using Routine Care Data

Hendrikus J. A. van Os, Jos P. Kanning, Tobias N. Bonten, Margot M. Rakers, Hein Putter, Mattijs E. Numans, Ynte M. Ruigrok, Rolf H. H. Groenwold, Marieke J. H. Wermer

Summary: This study aimed to develop prediction models for the risk of first-ever cardiovascular events in men and women aged 30 to 49 years. The study found that the models based on traditional cardiovascular predictors had moderate discriminative performance in terms of gender differences. However, the addition of nontraditional cardiovascular predictors improved the performance of the models.

JOURNAL OF THE AMERICAN HEART ASSOCIATION (2023)

Add to Collection

Editorial Material Ethics

Enabling the Nonhypothesis-Driven Approach: On Data Minimalization, Bias, and the Integration of Data Science in Medical Research and Practice

C. W. Safarlou, M. van Smeden, R. Vermeulen, K. R. Jongsma

AMERICAN JOURNAL OF BIOETHICS (2023)

Add to Collection

Article Mathematical & Computational Biology

The marginality principle revisited: Should higher-order terms always be accompanied by lower-order terms in regression analyses?

Tim P. Morris, Maarten van Smeden, Tra My Pham

Summary: The marginality principle guides analysts to include lower-order terms in models to avoid omitting higher-order terms. However, the determination of lower-order terms depends on the scale of measurement and is subjective.

BIOMETRICAL JOURNAL (2023)

Add to Collection

Article Cardiac & Cardiovascular Systems

Diagnostic management of acute pulmonary embolism: a prediction model based on a patient data meta-analysis

Nick van Es, Toshihiko Takada, Noemie Kraaijpoel, Frederikus A. Klok, Milou A. M. Stals, Harry R. Buller, D. Mark Courtney, Yonathan Freund, Javier Galipienzo, Gregoire Le Gal, Waleed Ghanima, Menno Huisman, Jeffrey A. Kline, Karel G. M. Moons, Sameer Parpia, Arnaud Perrier, Marc Righini, Helia Robert-Ebadi, Pierre-Marie Roy, Phil S. Wells, Kerstin de Wit, Maarten van Smeden, Geert-Jan Geersing

Summary: The aim of this study was to develop a clinical prediction model for acute pulmonary embolism (PE) based on readily available clinical items and D-dimer concentrations. The model showed good discrimination and calibration, and outperformed the currently widely-used Wells score and D-dimer testing methods.

EUROPEAN HEART JOURNAL (2023)

Add to Collection

Review Health Care Sciences & Services

Systematic review finds risk of bias and applicability concerns for models predicting central line-associated bloodstream infection

Shan Gao, Elena Albu, Krizia Tuand, Veerle Cossey, Frank Rademakers, Ben Van Calster, Laure Wynants

Summary: This systematic review assessed the risk of bias and applicability of published prediction models for central line-associated bloodstream infection (CLA-BSI) in hospitalized patients. The results revealed a lack of prediction models with potential clinical use and highlighted the urgent need for the development of an applicable model for CLA-BSI.

JOURNAL OF CLINICAL EPIDEMIOLOGY (2023)

Add to Collection

Letter Multidisciplinary Sciences

Provide public access to ethics-approved study protocols

Ben Van Calster

NATURE (2023)

Add to Collection

Article Health Care Sciences & Services

Multiclass risk models for ovarian malignancy: an illustration of prediction uncertainty due to the choice of algorithm

Ashleigh Ledger, Jolien Ceusters, Lil Valentin, Antonia Testa, Caroline Van Holsbeke, Dorella Franchi, Tom Bourne, Wouter Froyman, Dirk Timmerman, Ben Van Calster

Summary: Assessing the malignancy risk of ovarian tumors is crucial for selecting appropriate management. In this study, six algorithms were compared to estimate the probabilities of different types of ovarian tumors. The models developed showed good performance, but individual probability estimates varied substantially.

BMC MEDICAL RESEARCH METHODOLOGY (2023)

Add to Collection

Article Obstetrics & Gynecology

A blended preconception lifestyle programme for couples undergoing IVF: lessons learned from a multicentre randomized controlled trial

Tessy Boedt, Eline Dancet, Diane De Neubourg, Sofie Vereeck, Jan Seghers, Katleen van der Gucht, Ben Van Calster, Carl Spiessens, Sharon Lie Fong, Christophe Matthys

Summary: This randomized controlled trial aimed to assess the impact of a blended preconception lifestyle programme on reproductive and lifestyle outcomes of couples undergoing IVF. However, the trial was prematurely stopped due to the Covid-19 pandemic and the results indicate that the programme did not have a meaningful effect on time to ongoing pregnancy or other reproductive and lifestyle outcomes.

HUMAN REPRODUCTION OPEN (2023)

Add to Collection

No Data Available

© Peeref 2019-2024. All rights reserved.