Scholarlink

Background <p>Viruses represent the most numerous’biological entities’on Earth; but the direct quantification of viruses within ecosystems is a significant challenge. The classical method of epifluorescence microscopy (EFM) remains the gold standard measurement of viral-like particles (VLPs) within ecosystems. Quantifying VLPs in epifluorescence microscopy is burdened by ongoing challenges that include manual human counting, an absence of accurate morphological sizing, and the range of viral sizes falling below the diffraction limit of light microscopy.</p> Methods <p>EpiVirQuant utilizes epifluorescence microscopy images of VLPs from environmental and/or clinical samples then quantifies VLP particle size and abundance using computational image analysis. The software analyses VLP images in four steps: (i) optimization box selection, (ii) blind deconvolution, (iii) derivation of a size-correction factor, and (iv) enumeration and sizing of VLPs. Sizing occurs by the spiked-in of DAPI beads of a known size via a calibrate image and by deriving a tunable point-spread function (tPSF).</p> Results <p>Due to microscope and user variation the calibration image choice substantially impacts the derived tPSF and resulting size estimates. This image should be very high quality, lacking blur, with clear DAPI beads and VLPs present. tPSF was benchmarked against commonly used blind-deconvolution PSFs (e.g. Gaussian PSF). tPSF outperformed other PSFs converging to the target microsphere diameter fastest while maintaining stability over increasing Richardson–Lucy iterations. Using the selected tPSF and calibration image, automated counts from EpiVirQuant closely matched expert human counts, with only ~2% more VLPs identified overall by EpiVirQuant. Mean VLP diameter was ~175 nm, consistent with reported sizes for VLPs in hypersaline environments. The tPSF effectively resolved closely spaced particles that otherwise appeared merged, improving both enumeration and size accuracy. Outlier objects ( &gt; 500 nm) and candidates with extreme or ambiguous linear eccentricity were conservatively removed.</p> Discussion <p>Here, a proof-of-concept computer vision framework for the automated enumeration and sizing of viral-like particles is presented, known as EpiVirQuant. A novel tunable point spread function is introduced which allows for a dynamic blind deconvolution. EpiVirQuant quantified VLPs in the 140–210 nm range which is consistent for VLPs within aquatic ecosystems. Runtime ranged from 60–80 seconds per image depending on parameter selection. This provides a viable proof-of-concept cost-effective solution for the enumeration and large-scale morphological analysis of VLPs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Resolving and quantifying viral-like particles via blind deconvolution

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月9日

查看全文
  • Jose L. Figueroa III,
  • Sadie M. Hollenack,
  • Madeline Bellanger-Perry,
  • Bryan Fulghum,
  • Pieter T. Visscher,
  • Richard Allen White III

摘要

Background

Viruses represent the most numerous’biological entities’on Earth; but the direct quantification of viruses within ecosystems is a significant challenge. The classical method of epifluorescence microscopy (EFM) remains the gold standard measurement of viral-like particles (VLPs) within ecosystems. Quantifying VLPs in epifluorescence microscopy is burdened by ongoing challenges that include manual human counting, an absence of accurate morphological sizing, and the range of viral sizes falling below the diffraction limit of light microscopy.

Methods

EpiVirQuant utilizes epifluorescence microscopy images of VLPs from environmental and/or clinical samples then quantifies VLP particle size and abundance using computational image analysis. The software analyses VLP images in four steps: (i) optimization box selection, (ii) blind deconvolution, (iii) derivation of a size-correction factor, and (iv) enumeration and sizing of VLPs. Sizing occurs by the spiked-in of DAPI beads of a known size via a calibrate image and by deriving a tunable point-spread function (tPSF).

Results

Due to microscope and user variation the calibration image choice substantially impacts the derived tPSF and resulting size estimates. This image should be very high quality, lacking blur, with clear DAPI beads and VLPs present. tPSF was benchmarked against commonly used blind-deconvolution PSFs (e.g. Gaussian PSF). tPSF outperformed other PSFs converging to the target microsphere diameter fastest while maintaining stability over increasing Richardson–Lucy iterations. Using the selected tPSF and calibration image, automated counts from EpiVirQuant closely matched expert human counts, with only ~2% more VLPs identified overall by EpiVirQuant. Mean VLP diameter was ~175 nm, consistent with reported sizes for VLPs in hypersaline environments. The tPSF effectively resolved closely spaced particles that otherwise appeared merged, improving both enumeration and size accuracy. Outlier objects ( > 500 nm) and candidates with extreme or ambiguous linear eccentricity were conservatively removed.

Discussion

Here, a proof-of-concept computer vision framework for the automated enumeration and sizing of viral-like particles is presented, known as EpiVirQuant. A novel tunable point spread function is introduced which allows for a dynamic blind deconvolution. EpiVirQuant quantified VLPs in the 140–210 nm range which is consistent for VLPs within aquatic ecosystems. Runtime ranged from 60–80 seconds per image depending on parameter selection. This provides a viable proof-of-concept cost-effective solution for the enumeration and large-scale morphological analysis of VLPs.