Implementing Trust in Non-Small Cell Lung Cancer Diagnosis with a Conformalized Uncertainty-Aware AI Framework in Whole-Slide Images

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 30, 2024

Ensuring trustworthiness is fundamental to the development of artificial intelligence (AI) that considered societally responsible, particularly in cancer diagnostics, where a misdiagnosis can have dire consequences. Current digital pathology AI models lack systematic solutions address concerns arising from model limitations and data discrepancies between deployment environments. To this issue, we developed TRUECAM, framework designed ensure both non-small cell lung subtyping with whole-slide images. TRUECAM integrates 1) spectral-normalized neural Gaussian process for identifying out-of-scope inputs 2) an ambiguity-guided elimination tiles filter out highly ambiguous regions, addressing trustworthiness, as well 3) conformal prediction controlled error rates. We systematically evaluated across multiple large-scale datasets, leveraging task-specific foundation models, illustrate wrapped significantly outperforms such guidance, terms classification accuracy, robustness, interpretability, efficiency, while also achieving improvements fairness. These findings highlight versatile wrapper diverse architectural designs, promoting their responsible effective applications real-world settings.

Language: Английский

Functionalized Multichannel Fluorescence-Encoded Nanosystem on Erythrocyte-Coated Nanoparticles for Precise Cancer Subtype Discrimination

Nano Letters, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 17, 2024

Rapid and precise cancer subtype discrimination is essential for personalized oncology. Conventional diagnostic methods often lack sufficient accuracy speed. Here, we introduce a multichannel fluorescence-encoded nanosystem based on erythrocyte-coated polydopamine nanoparticles (PDA@EM), functionalized with multiple resurfaced fluorescent proteins. The fluorescence of these proteins initially quenched by PDA@EM restored upon cell addition. This enables highly sensitive "turn-on" profiling cells within 30 min, achieving 100% in distinguishing various classifying wide range lines, including subtypes oral squamous carcinoma (OSCC). Notably, it offers rapid, label-free diagnostics OSCC malignancy from clinical samples postsurgery. capability was validated through histopathological proteomic analyses, which identified protein signatures associated tumor progression immune suppression. Overall, our nanosensor represents an advanced molecular platform, paving the way treatment

Language: Английский