Why we should not mistake accuracy of medical AI for efficiency.
Ontology highlight
ABSTRACT:
SUBMITTER: Jongsma KR
PROVIDER: S-EPMC10912629 | biostudies-literature | 2024 Mar
REPOSITORIES: biostudies-literature
Publications
Why we should not mistake accuracy of medical AI for efficiency.
Jongsma Karin Rolanda KR Sand Martin M Milota Megan M
NPJ digital medicine 20240304 1
Similar Datasets
Project description: Not available
| S-EPMC10787826 | biostudies-literature
Project description:BackgroundArtificial intelligence (AI) models are increasingly used in the medical domain. However, as medical data is highly sensitive, special precautions to ensure its protection are required. The gold standard for privacy preservation is the introduction of differential privacy (DP) to model training. Prior work indicates that DP has negative implications on model accuracy and fairness, which are unacceptable in medicine and represent a main barrier to the widespread use of privacy-preserving techniques. In this work, we evaluated the effect of privacy-preserving training of AI models regarding accuracy and fairness compared to non-private training.MethodsWe used two datasets: (1) A large dataset (N = 193,311) of high quality clinical chest radiographs, and (2) a dataset (N = 1625) of 3D abdominal computed tomography (CT) images, with the task of classifying the presence of pancreatic ductal adenocarcinoma (PDAC). Both were retrospectively collected and manually labeled by experienced radiologists. We then compared non-private deep convolutional neural networks (CNNs) and privacy-preserving (DP) models with respect to privacy-utility trade-offs measured as area under the receiver operating characteristic curve (AUROC), and privacy-fairness trade-offs, measured as Pearson's r or Statistical Parity Difference.ResultsWe find that, while the privacy-preserving training yields lower accuracy, it largely does not amplify discrimination against age, sex or co-morbidity. However, we find an indication that difficult diagnoses and subgroups suffer stronger performance hits in private training.ConclusionsOur study shows that - under the challenging realistic circumstances of a real-life clinical dataset - the privacy-preserving training of diagnostic deep learning models is possible with excellent diagnostic accuracy and fairness.
| S-EPMC10940659 | biostudies-literature
Project description: Not available
| S-EPMC11322186 | biostudies-literature
Project description:The prime editing (PE) system consists of a Cas9 nickase fused to a reverse transcriptase, which introduces precise edits into the target genomic region guided by a prime editing guide RNA. However, PE efficiency is limited by mismatch repair. To overcome this limitation, transient expression of a dominant-negative MLH1 (MLH1dn) has been used to inhibit key components of mismatch repair. Here, we designed a de novo MLH1 small binder (MLH1-SB) that binds to the dimeric interface of MLH1 and PMS2 using RFdiffusion and AlphaFold 3. The compact size of MLH1-SB enabled its integration into existing PE architectures via 2A systems, creating a novel PE-SB platform. The PE7-SB system significantly improved PE efficiency, achieving an 18.8-fold increase over PEmax and a 2.5-fold increase over PE7 in HeLa cells, as well as a 3.4-fold increase over PE7 in mice. This study highlights the potential of generative AI in advancing genome editing technology.
2025-04-07 | GSE293633 | GEO
Project description: Not available
| S-EPMC11322447 | biostudies-literature
Project description:This study evaluates multimodal AI models' accuracy and responsiveness in answering NEJM Image Challenge questions, juxtaposed with human collective intelligence, underscoring AI's potential and current limitations in clinical diagnostics. Anthropic's Claude 3 family demonstrated the highest accuracy among the evaluated AI models, surpassing the average human accuracy, while collective human decision-making outperformed all AI models. GPT-4 Vision Preview exhibited selectivity, responding more to easier questions with smaller images and longer questions.
| S-EPMC11306783 | biostudies-literature
Project description:Background and objectiveMedical image segmentation is a vital aspect of medical image processing, allowing healthcare professionals to conduct precise and comprehensive lesion analyses. Traditional segmentation methods are often labor intensive and influenced by the subjectivity of individual physicians. The advent of artificial intelligence (AI) has transformed this field by reducing the workload of physicians, and improving the accuracy and efficiency of disease diagnosis. However, conventional AI techniques are not without challenges. Issues such as inexplicability, uncontrollable decision-making processes, and unpredictability can lead to confusion and uncertainty in clinical decision-making. This review explores the evolution of AI in medical image segmentation, focusing on the development and impact of explainable AI (XAI) and trustworthy AI (TAI).MethodsThis review synthesizes existing literature on traditional segmentation methods, AI-based approaches, and the transition from conventional AI to XAI and TAI. The review highlights the key principles and advancements in XAI that aim to address the shortcomings of conventional AI by enhancing transparency and interpretability. It further examines how TAI builds on XAI to improve the reliability, safety, and accountability of AI systems in medical image segmentation.Key content and findingsXAI has emerged as a solution to the limitations of conventional AI by providing greater transparency and interpretability, allowing healthcare professionals to better understand and trust AI-driven decisions. However, XAI itself faces challenges, including those related to safety, robustness, and value alignment. TAI has been developed to overcome these challenges, offering a more reliable framework for AI applications in medical image segmentation. By integrating the principles of XAI with enhanced safety and dependability, TAI addresses the critical need for TAI systems in clinical settings.ConclusionsTAI presents a promising future for medical image segmentation, combining the benefits of AI with improved reliability and safety. Thus, TAI is a more viable and dependable option for healthcare applications, and could ultimately lead to better clinical outcomes for patients, and advance the field of medical image processing.
| S-EPMC11651983 | biostudies-literature
Project description:Proper functioning of biological cells requires that the process of protein expression be carried out with high efficiency and fidelity. Given an amino-acid sequence of a protein, multiple degrees of freedom still remain that may allow evolution to tune efficiency and fidelity for each gene under various conditions and cell types. Particularly, the redundancy of the genetic code allows the choice between alternative codons for the same amino acid, which, although 'synonymous,' may exert dramatic effects on the process of translation. Here we review modern developments in genomics and systems biology that have revolutionized our understanding of the multiple means by which translation is regulated. We suggest new means to model the process of translation in a richer framework that will incorporate information about gene sequences, the tRNA pool of the organism and the thermodynamic stability of the mRNA transcripts. A practical demonstration of a better understanding of the process would be a more accurate prediction of the proteome, given the transcriptome at a diversity of biological conditions.
| S-EPMC3101949 | biostudies-other
Project description:Interventions: Gold Standard:The diagnosis of the senior pathologist;Index test:The sensitivity and specificity of pathologists when working with and without AI assistance.
Primary outcome(s): Sensitivity;Specificity
Study Design: Diagnostic test for accuracy
| 113777 | ecrin-mdr-crc
Project description:Jitter-type spike resampling methods are routinely applied in neurophysiology for detecting temporal structure in spike trains (point processes). Several variations have been proposed. The concern has been raised, based on numerical experiments involving Poisson spike processes, that such procedures can be conservative. We study the issue and find it can be resolved by reemphasizing the distinction between spike-centered (basic) jitter and interval jitter. Focusing on spiking processes with no temporal structure, interval jitter generates an exact hypothesis test, guaranteeing valid conclusions. In contrast, such a guarantee is not available for spike-centered jitter. We construct explicit examples in which spike-centered jitter hallucinates temporal structure, in the sense of exaggerated false-positive rates. Finally, we illustrate numerically that Poisson approximations to jitter computations, while computationally efficient, can also result in inaccurate hypothesis tests. We highlight the value of classical statistical frameworks for guiding the design and interpretation of spike resampling methods.
| S-EPMC5955204 | biostudies-literature