Limits...
A contact-imaging based microfluidic cytometer with machine-learning for single-frame super-resolution processing.

Huang X, Guo J, Wang X, Yan M, Kang Y, Yu H - PLoS ONE (2014)

Bottom Line: This paper introduces a single-frame super-resolution processing with on-line machine-learning for contact images of cells.A corresponding contact-imaging based microfluidic cytometer prototype is demonstrated for cell recognition and counting.Compared with commercial flow cytometer, less than 8% error is observed for absolute number of microbeads; and 0.10 coefficient of variation is observed for cell-ratio of mixed RBC and HepG2 cells in solution.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore, Singapore.

ABSTRACT
Lensless microfluidic imaging with super-resolution processing has become a promising solution to miniaturize the conventional flow cytometer for point-of-care applications. The previous multi-frame super-resolution processing system can improve resolution but has limited cell flow rate and hence low throughput when capturing multiple subpixel-shifted cell images. This paper introduces a single-frame super-resolution processing with on-line machine-learning for contact images of cells. A corresponding contact-imaging based microfluidic cytometer prototype is demonstrated for cell recognition and counting. Compared with commercial flow cytometer, less than 8% error is observed for absolute number of microbeads; and 0.10 coefficient of variation is observed for cell-ratio of mixed RBC and HepG2 cells in solution.

Show MeSH

Related in: MedlinePlus

ELM-SR off-line training images for HepG2 and RBC cells.(A) The original HR images for HepG2 cell with two different appearances; and the same for RBC cells. (B) The corresponding LR images. (C) The interpolated images of LR images, which cannot show HF details. (D) The extracted HF components. The scale bar indicates 5 µm.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4128713&req=5

pone-0104539-g006: ELM-SR off-line training images for HepG2 and RBC cells.(A) The original HR images for HepG2 cell with two different appearances; and the same for RBC cells. (B) The corresponding LR images. (C) The interpolated images of LR images, which cannot show HF details. (D) The extracted HF components. The scale bar indicates 5 µm.

Mentions: The off-line training HR image library of HepG2 and RBC was first built. The raw HR images of HepG2 and RBC were taken by the microscope camera at ×40 objective, and saved into the HR image library with the size of 48×48, as shown in Fig. 6 (A1–A4). Then, the corresponding 12×12 LR images were obtained by bicubic down sampling the HR images as shown in Fig. 6 (B1–B4). Next, these LR cell images were interpolated back to the same size of their original HR images, i.e., 48×48. Note that the detailed structures cannot be observed from the interpolated images because the interpolation cannot recover the HF components, as shown in Fig. 6 (C1–C4).


A contact-imaging based microfluidic cytometer with machine-learning for single-frame super-resolution processing.

Huang X, Guo J, Wang X, Yan M, Kang Y, Yu H - PLoS ONE (2014)

ELM-SR off-line training images for HepG2 and RBC cells.(A) The original HR images for HepG2 cell with two different appearances; and the same for RBC cells. (B) The corresponding LR images. (C) The interpolated images of LR images, which cannot show HF details. (D) The extracted HF components. The scale bar indicates 5 µm.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4128713&req=5

pone-0104539-g006: ELM-SR off-line training images for HepG2 and RBC cells.(A) The original HR images for HepG2 cell with two different appearances; and the same for RBC cells. (B) The corresponding LR images. (C) The interpolated images of LR images, which cannot show HF details. (D) The extracted HF components. The scale bar indicates 5 µm.
Mentions: The off-line training HR image library of HepG2 and RBC was first built. The raw HR images of HepG2 and RBC were taken by the microscope camera at ×40 objective, and saved into the HR image library with the size of 48×48, as shown in Fig. 6 (A1–A4). Then, the corresponding 12×12 LR images were obtained by bicubic down sampling the HR images as shown in Fig. 6 (B1–B4). Next, these LR cell images were interpolated back to the same size of their original HR images, i.e., 48×48. Note that the detailed structures cannot be observed from the interpolated images because the interpolation cannot recover the HF components, as shown in Fig. 6 (C1–C4).

Bottom Line: This paper introduces a single-frame super-resolution processing with on-line machine-learning for contact images of cells.A corresponding contact-imaging based microfluidic cytometer prototype is demonstrated for cell recognition and counting.Compared with commercial flow cytometer, less than 8% error is observed for absolute number of microbeads; and 0.10 coefficient of variation is observed for cell-ratio of mixed RBC and HepG2 cells in solution.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore, Singapore.

ABSTRACT
Lensless microfluidic imaging with super-resolution processing has become a promising solution to miniaturize the conventional flow cytometer for point-of-care applications. The previous multi-frame super-resolution processing system can improve resolution but has limited cell flow rate and hence low throughput when capturing multiple subpixel-shifted cell images. This paper introduces a single-frame super-resolution processing with on-line machine-learning for contact images of cells. A corresponding contact-imaging based microfluidic cytometer prototype is demonstrated for cell recognition and counting. Compared with commercial flow cytometer, less than 8% error is observed for absolute number of microbeads; and 0.10 coefficient of variation is observed for cell-ratio of mixed RBC and HepG2 cells in solution.

Show MeSH
Related in: MedlinePlus