Computational Optics Enables Breast Cancer Profiling in Point-of-Care Settings

Abstract

The global burden of cancer, severe diagnostic bottlenecks in underserved regions, and underfunded health care systems are fueling the need for inexpensive, rapid, and treatment-informative diagnostics. On the basis of advances in computational optics and deep learning, we have developed a low-cost digital system, termed AIDA (artificial intelligence diffraction analysis), for breast cancer diagnosis of fine needle aspirates. Here, we show high accuracy (>90%) in (i) recognizing cells directly from diffraction patterns and (ii) classifying breast cancer types using deep-learning-based analysis of sample aspirates. The image algorithm is fast, enabling cellular analyses at high throughput (∼3 s per 1000 cells), and the unsupervised processing allows use by lower skill health care workers. AIDA can perform quantitative molecular profiling on individual cells, revealing intratumor molecular heterogeneity, and has the potential to improve cancer diagnosis and treatment. The system could be further developed for other cancers and thus find widespread use in global health.

Glass Chemistry to Analyze Human Cells under Adverse Conditions

“…Here, we show that glass slides can be activated, lyophilized, and used for immunocapture and subsequent analysis, all in kit format. As a part of NCI’s Center for Global Health and their Affordable Cancer Technologies program ( ), we are preparing to use the developed technology to perform clinical trials in Africa with previously developed analysis techniques for cell capture such as artificial intelligence diffraction analysis 4 and a low-cost contrast-enhanced micrography device that has been validated in a clinical trial for the molecular diagnosis of lymphoma. 1 The trials will be conducted in Botswana in both hospital and rural health center settings.…”

Nanomedicine of tyrosine kinase inhibitors

“…With growing number of public databases 151 and high-throughput methods, mining for information in large datasets becomes more and more important. The application of machine learning and AI technology introduces broad possibilities, such as drug discovery 152 , target discovery 153 , 154 , predictions of NP properties 150 , 155 , prediction of phenotypic responses 156 , 157 , image analysis for clinical diagnosis 158 , 159 , determination of suitable drug combinations for individualized therapy 160 , 161 with subsequent efficacy evaluation and dose adjustments 162 and so on 163 . Overall, computational methods and AI could be an integral part of processes vital to human medicine ( Figure 7 ).…”

New technology on the horizon: Fast analytical screening technique FNA (FAST‐FNA) enables rapid, multiplex biomarker analysis in head and neck cancers

“…A version currently in clinical trials is a deep‐learning–enabled fluorescence cytometer, which is a stand‐alone unit weighting approximately 6 pounds. Prototype versions of this instrumentation were originally developed for global health applications 18‐20 and are currently being adapted for high‐resolution multiplexed image cytometry. Figure 3 illustrates the FAST‐FNA pipeline technology for HNSCC, including FNAB sample collection and staining with FAST antibodies in cyclic fashion, image processing, the use of a deep‐learning algorithm, and the generation of quantitative biomarker data.…”