Clinical artificial intelligence quality improvement: towards continual monitoring and updating of AI algorithms in healthcare

npj Digital Medicine, Год журнала: 2022, Номер 5(1)

Опубликована: Май 31, 2022

Machine learning (ML) and artificial intelligence (AI) algorithms have the potential to derive insights from clinical data improve patient outcomes. However, these highly complex systems are sensitive changes in environment liable performance decay. Even after their successful integration into practice, ML/AI should be continuously monitored updated ensure long-term safety effectiveness. To bring AI maturity care, we advocate for creation of hospital units responsible quality assurance improvement algorithms, which refer as "AI-QI" units. We discuss how tools that long been used can adapted monitor static ML algorithms. On other hand, procedures continual model updating still nascent. highlight key considerations when choosing between existing methods opportunities methodological innovation.

Язык: Английский

---

Evaluation of clinical prediction models (part 1): from development to external validation

BMJ, Год журнала: 2024, Номер unknown, С. e074819 - e074819

Опубликована: Янв. 8, 2024

Evaluating the performance of a clinical prediction model is crucial to establish its predictive accuracy in populations and settings intended for use. In this article, first three part series, Collins colleagues describe importance meaningful evaluation using internal, internal-external, external validation, as well exploring heterogeneity, fairness, generalisability performance.

Язык: Английский

---

A systematic review and meta-analysis of artificial intelligence versus clinicians for skin cancer diagnosis

npj Digital Medicine, Год журнала: 2024, Номер 7(1)

Опубликована: Май 14, 2024

Scientific research of artificial intelligence (AI) in dermatology has increased exponentially. The objective this study was to perform a systematic review and meta-analysis evaluate the performance AI algorithms for skin cancer classification comparison clinicians with different levels expertise. Based on PRISMA guidelines, 3 electronic databases (PubMed, Embase, Cochrane Library) were screened relevant articles up August 2022. quality studies assessed using QUADAS-2. A sensitivity specificity performed accuracy clinicians. Fifty-three included review, 19 met inclusion criteria meta-analysis. Considering all subgroups clinicians, we found (Sn) (Sp) 87.0% 77.1% algorithms, respectively, Sn 79.78% Sp 73.6% (overall); differences statistically significant both Sp. difference between (Sn 92.5%, 66.5%) vs. generalists 64.6%, 72.8%), greater, when compared expert Performance 86.3%, 78.4%) vs dermatologists 84.2%, 74.4%) clinically comparable. Limitations clinical practice should be considered, future focus real-world settings, towards AI-assistance.

Язык: Английский

---

Auto-detection of the coronavirus disease by using deep convolutional neural networks and X-ray photographs

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Янв. 4, 2024

Abstract The most widely used method for detecting Coronavirus Disease 2019 (COVID-19) is real-time polymerase chain reaction. However, this has several drawbacks, including high cost, lengthy turnaround time results, and the potential false-negative results due to limited sensitivity. To address these issues, additional technologies such as computed tomography (CT) or X-rays have been employed diagnosing disease. Chest are more commonly than CT scans widespread availability of X-ray machines, lower ionizing radiation, cost equipment. COVID-19 presents certain radiological biomarkers that can be observed through chest X-rays, making it necessary radiologists manually search biomarkers. process time-consuming prone errors. Therefore, there a critical need develop an automated system evaluating X-rays. Deep learning techniques expedite process. In study, deep learning-based called Custom Convolutional Neural Network (Custom-CNN) proposed identifying infection in Custom-CNN model consists eight weighted layers utilizes strategies like dropout batch normalization enhance performance reduce overfitting. approach achieved classification accuracy 98.19% aims accurately classify COVID-19, normal, pneumonia samples.

Язык: Английский

---

Innovative applications of artificial intelligence during the COVID-19 pandemic

Infectious Medicine, Год журнала: 2024, Номер 3(1), С. 100095 - 100095

Опубликована: Фев. 21, 2024

The COVID-19 pandemic has created unprecedented challenges worldwide. Artificial intelligence (AI) technologies hold tremendous potential for tackling key aspects of management and response. In the present review, we discuss possibilities AI technology in addressing global posed by pandemic. First, outline multiple impacts current on public health, economy, society. Next, focus innovative applications advanced areas such as prediction, detection, control, drug discovery treatment. Specifically, AI-based predictive analytics models can use clinical, epidemiological, omics data to forecast disease spread patient outcomes. Additionally, deep neural networks enable rapid diagnosis through medical imaging. Intelligent systems support risk assessment, decision-making, social sensing, thereby improving epidemic control health policies. Furthermore, high-throughput virtual screening enables accelerate identification therapeutic candidates opportunities repurposing. Finally, future research directions combating COVID-19, emphasizing importance interdisciplinary collaboration. Though promising, barriers related model generalization, quality, infrastructure readiness, ethical risks must be addressed fully translate these innovations into real-world impacts. Multidisciplinary collaboration engaging diverse expertise stakeholders is imperative developing robust, responsible, human-centered solutions against emergencies.

Язык: Английский

---

FUTURE-AI: international consensus guideline for trustworthy and deployable artificial intelligence in healthcare

BMJ, Год журнала: 2025, Номер unknown, С. e081554 - e081554

Опубликована: Фев. 5, 2025

Despite major advances in artificial intelligence (AI) research for healthcare, the deployment and adoption of AI technologies remain limited clinical practice. This paper describes FUTURE-AI framework, which provides guidance development trustworthy tools healthcare. The Consortium was founded 2021 comprises 117 interdisciplinary experts from 50 countries representing all continents, including scientists, researchers, biomedical ethicists, social scientists. Over a two year period, guideline established through consensus based on six guiding principles—fairness, universality, traceability, usability, robustness, explainability. To operationalise set 30 best practices were defined, addressing technical, clinical, socioethical, legal dimensions. recommendations cover entire lifecycle healthcare AI, design, development, validation to regulation, deployment, monitoring.

Язык: Английский

---

Application of artificial intelligence in forecasting survival in high-grade glioma: systematic review and meta-analysis involving 79,638 participants

Neurosurgical Review, Год журнала: 2025, Номер 48(1)

Опубликована: Фев. 15, 2025

Язык: Английский

---

Quod erat demonstrandum? - Towards a typology of the concept of explanation for the design of explainable AI

Expert Systems with Applications, Год журнала: 2022, Номер 213, С. 118888 - 118888

Опубликована: Сен. 24, 2022

In this paper, we present a fundamental framework for defining different types of explanations AI systems and the criteria evaluating their quality. Starting from structural view how can be constructed, i.e., in terms an explanandum (what needs to explained), multiple explanantia (explanations, clues, or parts information that explain), relationship linking explanantia, propose explanandum-based typology point other possible typologies based on are presented they relate explanandia. We also highlight two broad complementary perspectives quality assessing explainability: epistemological psychological (cognitive). These definition attempts aim support three main functions believe should attract interest further research XAI scholars: clear inventories, verification criteria, validation methods.

Язык: Английский

---

Targeted validation: validating clinical prediction models in their intended population and setting

Diagnostic and Prognostic Research, Год журнала: 2022, Номер 6(1)

Опубликована: Дек. 22, 2022

Abstract Clinical prediction models must be appropriately validated before they can used. While validation studies are sometimes carefully designed to match an intended population/setting of the model, it is common for take place with arbitrary datasets, chosen convenience rather than relevance. We call estimating how well a model performs within “targeted validation”. Use this term sharpens focus on use which may increase applicability developed models, avoid misleading conclusions, and reduce research waste. It also exposes that external not required when population matches used develop model; here, robust internal sufficient, especially if development dataset was large.

Язык: Английский

---

Recommendations on compiling test datasets for evaluating artificial intelligence solutions in pathology

Modern Pathology, Год журнала: 2022, Номер 35(12), С. 1759 - 1769

Опубликована: Сен. 10, 2022

Artificial intelligence (AI) solutions that automatically extract information from digital histology images have shown great promise for improving pathological diagnosis. Prior to routine use, it is important evaluate their predictive performance and obtain regulatory approval. This assessment requires appropriate test datasets. However, compiling such datasets challenging specific recommendations are missing. A committee of various stakeholders, including commercial AI developers, pathologists, researchers, discussed key aspects conducted extensive literature reviews on in pathology. Here, we summarize the results derive general We address several questions: Which how many needed? How deal with low-prevalence subsets? can potential bias be detected? should reported? What requirements different countries? The intended help developers demonstrate utility products pathologists agencies verify reported measures. Further research needed formulate criteria sufficiently representative so operate less user intervention better support diagnostic workflows future.

Язык: Английский

---