Towards equitable AI in oncology

Nature Reviews Clinical Oncology, Год журнала: 2024, Номер 21(8), С. 628 - 637

Опубликована: Июнь 7, 2024

Язык: Английский

---

Artificial intelligence-based biomarkers for treatment decisions in oncology

Trends in cancer, Год журнала: 2025, Номер unknown

Опубликована: Янв. 1, 2025

The development of new therapeutic strategies such as immune checkpoint inhibitors (ICIs) and targeted therapies has increased the complexity treatment landscape for solid tumors. At current rate annual FDA approvals, potential options could increase by tenfold over next 5 years. cost personalized medicine technologies limits its accessibility, thus increasing socioeconomic disparities in treated population. In this review we describe artificial intelligence (AI)-based solutions - including deep learning (DL) methods routine medical imaging large language models (LLMs) electronic health records (EHRs) to support cancer decisions with cost-effective biomarkers. We address limitations these propose steps towards their adoption clinical practice.

Язык: Английский

---

Building an Ethical and Trustworthy Biomedical AI Ecosystem for the Translational and Clinical Integration of Foundation Models

Bioengineering, Год журнала: 2024, Номер 11(10), С. 984 - 984

Опубликована: Сен. 29, 2024

Foundation Models (FMs) are gaining increasing attention in the biomedical artificial intelligence (AI) ecosystem due to their ability represent and contextualize multimodal data. These capabilities make FMs a valuable tool for variety of tasks, including reasoning, hypothesis generation, interpreting complex imaging In this review paper, we address unique challenges associated with establishing an ethical trustworthy AI ecosystem, particular focus on development downstream applications. We explore strategies that can be implemented throughout pipeline effectively tackle these challenges, ensuring translated responsibly into clinical translational settings. Additionally, emphasize importance key stewardship co-design principles not only ensure robust regulation but also guarantee interests all stakeholders—especially those involved or affected by applications—are adequately represented. aim empower community harness models effectively. As navigate exciting frontier, our collective commitment stewardship, co-design, responsible translation will instrumental evolution truly enhances patient care medical decision-making, ultimately leading more equitable ecosystem.

Язык: Английский

---

AI in Histopathology Explorer for comprehensive analysis of the evolving AI landscape in histopathology

npj Digital Medicine, Год журнала: 2025, Номер 8(1)

Опубликована: Март 12, 2025

Abstract Digital pathology and artificial intelligence (AI) hold immense transformative potential to revolutionize cancer diagnostics, treatment outcomes, biomarker discovery. Gaining a deeper understanding of deep learning algorithm methods applied histopathological data evaluating their performance on different tasks is crucial for developing the next generation AI technologies. To this end, we developed in Histopathology Explorer (HistoPathExplorer); an interactive dashboard with intelligent tools available at www.histopathexpo.ai . This real-time online resource enables users, including researchers, decision-makers, various stakeholders, assess current landscape applications specific clinical tasks, analyze performance, explore factors influencing translation into practice. Moreover, quality index was defined comprehensiveness methodological details published methods. HistoPathExplorer highlights opportunities challenges histopathology, offers valuable creating more effective shaping strategies guidelines translating digital

Язык: Английский

---

The clinical application of artificial intelligence in cancer precision treatment

Journal of Translational Medicine, Год журнала: 2025, Номер 23(1)

Опубликована: Янв. 27, 2025

Язык: Английский

---

Current Advancements in Digital Neuropathology and Machine Learning for the Study of Neurodegenerative Diseases

American Journal Of Pathology, Год журнала: 2025, Номер unknown

Опубликована: Фев. 1, 2025

Язык: Английский

---

Computational pathology: A comprehensive review of recent developments in digital and intelligent pathology

Опубликована: Март 1, 2025

Язык: Английский

---

Demographic bias of expert-level vision-language foundation models in medical imaging

Science Advances, Год журнала: 2025, Номер 11(13)

Опубликована: Март 26, 2025

Advances in artificial intelligence (AI) have achieved expert-level performance medical imaging applications. Notably, self-supervised vision-language foundation models can detect a broad spectrum of pathologies without relying on explicit training annotations. However, it is crucial to ensure that these AI do not mirror or amplify human biases, disadvantaging historically marginalized groups such as females Black patients. In this study, we investigate the algorithmic fairness state-of-the-art chest x-ray diagnosis across five globally sourced datasets. Our findings reveal compared board-certified radiologists, consistently underdiagnose groups, with even higher rates seen intersectional subgroups female Such biases present over wide range and demographic attributes. Further analysis model embedding uncovers its substantial encoding information. Deploying systems intensify preexisting care disparities, posing potential challenges equitable healthcare access raising ethical questions about their clinical

Язык: Английский

---

Computational pathology for breast cancer: where do we stand for prognostic applications?

The Breast, Год журнала: 2025, Номер unknown, С. 104464 - 104464

Опубликована: Март 1, 2025

The very early days of artificial intelligence (AI) in healthcare are behind us. AI is now spreading the sector and gradually being implemented routine clinical practice. Driven by increasing digitization microscope slides, computational pathology (CPath) further strengthening role field oncology. CPath transforming fundamental research as well practice, both for diagnostic prognostic applications. In breast cancer, holds potential to address several unmet needs, particularly areas biomarkers tools. Indeed, multiple applications on their way, ranging from predicting clinically meaningful endpoints offering alternatives gene-expression testing detecting molecular alterations directly digitized whole slide images. However, fully harness CPath, challenges must be overcome. These include improving availability multimodal patient data, advancing digitalization laboratories, adoption within medical community, navigating regulatory hurdles. This review offers an overview current landscape highlighting progress made hurdles that remain its widespread

Язык: Английский

---

NLP-enriched social determinants of health improve prediction of suicide death among the Veterans

Research Square (Research Square), Год журнала: 2025, Номер unknown

Опубликована: Март 31, 2025

Predictions of suicide death patients discharged from psychiatric hospitals (PDPH) can guide intervention efforts including intensive post-discharge case management programs, designed to reduce risk among high-risk patients. This study aims determine if additions social and behavioral determinants health (SBDH) as predictors could improve the prediction PDPH. We analyzed a cohort 197,581 US Veterans 129 VHA across between January 1, 2017, July 2019 with total 414,043 discharges. Predictive variables included administrative data SBDH, latter derived unstructured clinical notes via natural language processing (NLP) system ICD codes, observed within 365-day window prior discharge. evaluated impact SBDH on predictive performance two advanced models: an ensemble traditional machine learning models transformer-based deep foundation model for electronic records (TransformEHR). measured sensitivity, positive value (PPV), area under receiver operating characteristic curve (AUROC) overall by gender. Calibration analysis was also conducted measure reliability. TransformEHR achieved AUROC 64.04. Specifically, ICD-based improved 3.1% (95% CI, 1.6% – 4.5%) 2.9% 0.5% 5.4%) TransformEHR, compared without SBDH. NLP-extracted further AUROC: 1.7% 0.1%– 3.3%) 1.8% 0.6%– 2.9%) TransformEHR. 0.2%, 0.4%, 0.8%, PPV per 100 PDPH 7, 30, 90, 180 respectively. Moreover, showed superior calibration fairness model, improving both models. In conclusion, performance, calibration, after their

Язык: Английский

---