Lens-free smart microscope diagnoses meningitis
- 著者:Ella Cai
- 公開：:2017-07-18
Together with medical equipment firm Horiba, French research lab Leti has created a automated diagnosis system for meningitis, based around a simple lens-free microscope and a tablet computer.
Samples are introduced just above the sensor, with the arrangement causing holographic diffraction patterns to fall upon the sensor.
Computer processing of the diffraction patterns allows an image of the sample to be reconstructed. Resolution is approximately equal to the pixel size in X and Y, and 500nm in Z.
meningitis lens free microscopeBoth magnitude of light – as would be seen down an ordinary microscope – and the phase of the light can be extracted.
Importantly, the field of view – the area in focus – is approximately the whole image sensor size – 4x3mm in this case – so there is no need to move the sample around over the sensor.
Meningitis has been chosen as a target for diagnosis because it is detected by counting the number of white blood cells (‘leukocytes’, which are around 20µm across) in one microlitre of cerebrospinal fluid – which is tricky as there are thousands of other similar-sized particles legitimately floating round in this fluid.
Generally, and expert human looks down the microscope and takes up to 20 minutes (in a difficult case) to count the white blood cells, a representative of Horiba told Electronics Weekly, adding that its research indicated a 20% variability in count across five experts (see reference below for exact figures).
There are no white blood cells in the fluid in a healthy human, and >10 cells/µl is an agreed diagnosis threshold.
Horiba’s system (see picture) combines the lens-free microscope and a 3µl sample chamber with a tablet computer.
To reconstruct both the magnitude and phase images from the raw diffraction pattern, classify all the particles into white blood cell or not, and to count the white blood cells, takes 1-2 minutes on the tablet processor.
The classification algorithms are biased to produce no false negative diagnoses – it therefore catches 100% of cases. As a consequence it produces some false positives, with a specificity of 79% (so 21% of healthy people were said to have the illness).
What the demonstrator does not do is prepare the sample for the microscope.
Horiba already makes lab equipment that automatically prepares samples of body fluid and makes measurements, and Leti has expertise in micro-fluidics.
According to the Horiba spokeswoman, the company aims to combine micro-fluidic sample preparation with the demonstrated analysis technique to create point-of-care devices that can replace sending samples to a laboratory.
The lens-free microscope is physically simple (see demonstrator photo) and consists of a light source and a pin-hole, both mounted ~50mm above an image sensor – in this case a re-purposed 10Mpixel phone image sensor with 1.6µm pixels.
Samples are introduced just above the sensor, with the arrangement causing holographic diffraction patterns to fall upon the sensor.
Computer processing of the diffraction patterns allows an image of the sample to be reconstructed. Resolution is approximately equal to the pixel size in X and Y, and 500nm in Z.
meningitis lens free microscopeBoth magnitude of light – as would be seen down an ordinary microscope – and the phase of the light can be extracted.
Importantly, the field of view – the area in focus – is approximately the whole image sensor size – 4x3mm in this case – so there is no need to move the sample around over the sensor.
Meningitis has been chosen as a target for diagnosis because it is detected by counting the number of white blood cells (‘leukocytes’, which are around 20µm across) in one microlitre of cerebrospinal fluid – which is tricky as there are thousands of other similar-sized particles legitimately floating round in this fluid.
Generally, and expert human looks down the microscope and takes up to 20 minutes (in a difficult case) to count the white blood cells, a representative of Horiba told Electronics Weekly, adding that its research indicated a 20% variability in count across five experts (see reference below for exact figures).
There are no white blood cells in the fluid in a healthy human, and >10 cells/µl is an agreed diagnosis threshold.
Horiba’s system (see picture) combines the lens-free microscope and a 3µl sample chamber with a tablet computer.
To reconstruct both the magnitude and phase images from the raw diffraction pattern, classify all the particles into white blood cell or not, and to count the white blood cells, takes 1-2 minutes on the tablet processor.
The classification algorithms are biased to produce no false negative diagnoses – it therefore catches 100% of cases. As a consequence it produces some false positives, with a specificity of 79% (so 21% of healthy people were said to have the illness).
What the demonstrator does not do is prepare the sample for the microscope.
Horiba already makes lab equipment that automatically prepares samples of body fluid and makes measurements, and Leti has expertise in micro-fluidics.
According to the Horiba spokeswoman, the company aims to combine micro-fluidic sample preparation with the demonstrated analysis technique to create point-of-care devices that can replace sending samples to a laboratory.
