A novel method based on a Mask R-CNN model for processing dPCR images

Hu, Zhenming; Fang, Weibo; Gou, Tong; Wu, Wenrui; Hu, Jiumei; Zhou, Shufang; Mu, Ying

doi:10.1039/c9ay01005j

Cited by 38 publications

(33 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These challenges make the conventional image processing method with the thresholding and segmentation (GTS) scheme ( Figure S4A ) inaccurate and require human supervision for error correction in handling digital assay images. Mu et al ( Gou et al 2019 ; Hu et al 2019 ) showed that the use of machine learning algorithms would provide promising solutions to significantly improve the accuracy of digital assay image processing. However, this approach is only applied for single-color images with a small number of microreactors (a few thousand) with 1080×1120 pixels, which is impractical for high-throughput analysis.…”

Section: Resultsmentioning

confidence: 99%

Machine learning-based cytokine microarray digital immunoassay analysis

Song

Zhao

Cai

et al. 2021

Biosensors and Bioelectronics

View full text Add to dashboard Cite

Serial measurement of a large panel of protein biomarkers near the bedside could provide a promising pathway to transform the critical care of acutely ill patients. However, attaining the combination of high sensitivity and multiplexity with a short assay turnaround poses a formidable technological challenge. Here, the authors develop a rapid, accurate, and highly multiplexed microfluidic digital immunoassay by incorporating machine learning-based autonomous image analysis. The assay has achieved 12-plexed biomarker detection in sample volume < 15 μL at concentrations < 5pg/mL while only requiring a 5-min assay incubation, allowing for all processes from sampling to result to be completed within 40 min. The assay procedure applies both a spatial-spectral microfluidic encoding scheme and an image data analysis algorithm based on machine learning with a convolutional neural network (CNN) for pre-equilibrated single-molecule protein digital counting. This unique approach remarkably reduces errors facing the high-capacity multiplexing of digital immunoassay at low protein concentrations. Longitudinal data obtained for a panel of 12 serum cytokines in human patients receiving chimeric antigen receptor-T (CAR-T) cell therapy reveals the powerful biomarker profiling capability. The assay could also be deployed for near-real-time immune status monitoring of critically ill COVID-19 patients developing cytokine storm syndrome.

show abstract

Section: Resultsmentioning

confidence: 99%

Machine learning-based cytokine microarray digital immunoassay analysis

Song

Zhao

Cai

et al. 2021

Biosensors and Bioelectronics

View full text Add to dashboard Cite

show abstract

“…This gave superior performance. Mask R-CNN was initially applied to computer vision datasets such as common objects in context COCO [ 46 ] but has since been used in applications such as nuclei detection and recognition of target signals in digital polymerase chain reaction fluorescence images [ 64 ].…”

Section: Methodsmentioning

confidence: 99%

Developing Electron Microscopy Tools for Profiling Plasma Lipoproteins Using Methyl Cellulose Embedment, Machine Learning and Immunodetection of Apolipoprotein B and Apolipoprotein(a)

Giesecke

Soete

Mackinnon

et al. 2020

IJMS

View full text Add to dashboard Cite

Plasma lipoproteins are important carriers of cholesterol and have been linked strongly to cardiovascular disease (CVD). Our study aimed to achieve fine-grained measurements of lipoprotein subpopulations such as low-density lipoprotein (LDL), lipoprotein(a) (Lp(a), or remnant lipoproteins (RLP) using electron microscopy combined with machine learning tools from microliter samples of human plasma. In the reported method, lipoproteins were absorbed onto electron microscopy (EM) support films from diluted plasma and embedded in thin films of methyl cellulose (MC) containing mixed metal stains, providing intense edge contrast. The results show that LPs have a continuous frequency distribution of sizes, extending from LDL (> 15 nm) to intermediate density lipoprotein (IDL) and very low-density lipoproteins (VLDL). Furthermore, mixed metal staining produces striking “positive” contrast of specific antibodies attached to lipoproteins providing quantitative data on apolipoprotein(a)-positive Lp(a) or apolipoprotein B (ApoB)-positive particles. To enable automatic particle characterization, we also demonstrated efficient segmentation of lipoprotein particles using deep learning software characterized by a Mask Region-based Convolutional Neural Networks (R-CNN) architecture with transfer learning. In future, EM and machine learning could be combined with microarray deposition and automated imaging for higher throughput quantitation of lipoproteins associated with CVD risk.

show abstract

“…In contrast to a previously reported study (29), we greatly reduced the number of convolution layers and filters (depth of network) for high speed processing. Our algorithm employs much fewer labels and features required for imaging processing than those for other typical CNN applications, such as autonomous driving.…”

Section: Introductionmentioning

confidence: 98%

“…These challenges make the conventional image processing method with the thresholding and segmentation (GTS) scheme ( Figure S4A) inaccurate, thus requiring human supervision for error correction in handling digital assay images. Mu et al (29,30) showed that the use of machine learning algorithms would provide promising solutions to significantly improve the accuracy of digital assay image processing. However, this approach is only applied for single-color images with a small number of microreactors (a few thousand) with 1080×1120 pixels, which is impractical for high-throughput analysis.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning-Based Protein Microarray Digital Assay Analysis

Song

Zhao

Cai

et al. 2020

Preprint

View full text Add to dashboard Cite

Serial measurement of a large panel of protein biomarkers near the bedside could provide a promising pathway to transform the critical care of acutely ill patients. However, attaining the combination of high sensitivity and multiplexity with a short assay turnaround poses a formidable technological challenge. Here, we developed a rapid, accurate, and highly multiplexed microfluidic digital immunoassay by incorporating machine learning-based autonomous image analysis. The assay achieved 14-plexed biomarker detection at concentrations < 10pg/mL with a sample volume < 10 μL, including all processes from sampling to analyzed data delivery within 30 min, while only requiring a 5-min assay incubation. The assay procedure applied both a spatial-spectral microfluidic encoding scheme and an image data analysis algorithm based on machine learning with a convolutional neural network (CNN) for pre-equilibrated single-molecule protein digital counting. This unique approach remarkably reduced errors facing the high-capacity multiplexing of digital immunoassay at low protein concentrations. Longitudinal data obtained for a panel of 14 serum cytokines in human patients receiving chimeric antigen receptor-T (CAR-T) cell therapy manifested the powerful biomarker profiling capability and great potential of the assay for its translation to near-real-time bedside immune status monitoring.

show abstract

A novel method based on a Mask R-CNN model for processing dPCR images

Abstract: The flow of Mask R-CNN model for processing digital polymerase chain reaction (dPCR) fluorescence images.

Cited by 38 publications

References 32 publications

Machine learning-based cytokine microarray digital immunoassay analysis

Machine learning-based cytokine microarray digital immunoassay analysis

Developing Electron Microscopy Tools for Profiling Plasma Lipoproteins Using Methyl Cellulose Embedment, Machine Learning and Immunodetection of Apolipoprotein B and Apolipoprotein(a)

Machine Learning-Based Protein Microarray Digital Assay Analysis

Contact Info

Product

Resources

About