In this interview, Prof Bart Nicolaï shares his insights into the potential for advanced technologies to help research and detect postharvest disorders. By Anna Mouton
Nicolaï heads the Postharvest Research Group of the Biosystems Department at the University of Leuven in Belgium, and is the director of the Flanders Centre of Postharvest Technology, a public-private partnership between the University of Leuven and the Association of Belgian Horticultural Co-operatives.
Q What technology is currently used for in-line scanning in packhouses?
A The only technology that I know of is based on near-infrared measurements. The idea is that you irradiate the fruit with near-infrared light. The light interacts with the chemicals and microstructural environment in the fruit, and if you then look at how the light changes while it travels through the fruit you can learn something about what is inside.Read More
The typical application that has been developed since the 90s is the measurement of sugar content. But you can also get information about the microstructure which means that, for example, if you have internal disorders like browning you can pick that up. The challenge is that the penetration depth of near-infrared light is shallow, so you need to have a very powerful light source. And it is a point measurement, so if the disorder is not in the path of the light, then you might miss it.
With X-ray technologies you don't have that problem because you inspect the whole fruit. So I think it is a more viable technology for detecting internal disorders.
Q How did you start using X-ray scanning to study postharvest disorders?
A We were looking at controlled atmosphere storage of pears and apples, and we wanted to understand how internal browning develops. We approached this by constructing mathematical gas-transport models to test different scenarios on the computer, and to maybe come up with gas conditions to avoid browning disorders. In the process of constructing the models we decided to add information about the geometry of the fruit at different scales.
We look at three-dimensional microscopic images at a resolution of one micrometre and beyond. This is for our fundamental understanding of what happens in fruit during controlled atmosphere storage. We use basically the same technology as hospitals use for three-dimensional scanning but at a much higher resolution.
It then occurred to us that we could apply this technology in practice, for example to detect disorders in-line. Our two main research directions here are radiography and tomography.
Q What are the advantages and disadvantages of radiography and tomography?
A Radiographs are essentially the pictures you would get if you break your arm and go to a hospital. The image is fuzzy, but you can typically detect a clear bone structure. The problem with fruit is that they don't have bones – they don't have such a clear contrast.
So over recent years we have totally shifted to artificial intelligence to get information from these images. It is called deep learning – we train the software to recognise what we are looking for. We believe that this technology is almost ready for sorting-line companies to implement.
Tomography works really well in the sense that you have a lot of contrast in the images, and you can clearly see what and where the disorders are. But the disadvantage is that it is very hard to do this in-line. I don't think this is near to any commercial implementation in the next couple of years.
The first challenge is that figuring out the three-dimensional geometry from these two-dimensional images only works well when there's not too much motion. A second challenge is that you will only have images from some angles – you won't have a full image set. And you have to remember that these lines work at a speed of 10 – 15 fruit per second. And then there's little time to do the computations to reconstruct the three-dimensional image.
So we are looking at the minimum number of transmission images that are required. If we can detect a brown spot or a cavity, that's good enough. And the second approach is that we're also trying to use these deep learning techniques. The neural network may be able to work its way through the images and find the disorder without a full reconstruction.
In lab-scale and simulation conditions this seems to work quite well, but the challenge of course is always to implement this in a more practical environment.
Q Using this technology, have you found differences between newer and older cultivars?
A We have not specifically done work on this, but we do scan a lot of fruit. If you look at the most interesting new cultivars that have been introduced and are being introduced, like Jazz [cultivar Scifresh] or Kanzi [cultivar Nicoter], some of these novel cultivars are firmer than for example Golden Delicious.
It's reasonable to expect that there's a microstructural basis for this firmer texture. One aspect is probably the porosity of the fruit. In general, you would expect a softer texture in fruit with more pores. But there is also a purely biochemical component, and that is the composition of the cell walls, and also the expression of cell-wall enzymes.
For example, we compared cultivars such as Braeburn and Kanzi [cultivar Nicoter] to for example, Jonagold and there is a difference in the microstructure and also in porosity. But I'm not really sure yet whether the texture of these cultivars is due to differences in microstructure, or different cell-wall biochemistry.
These things are also some of the research questions that we are trying to investigate together with Elke [Dr Elke Crouch of the Department of Horticultural Sciences at Stellenbosch University].
THE MANY USES OF RADIATION
Visible light is the most familiar form of electromagnetic radiation because this is the light that humans can see. It is usually defined as light with a wavelength of 380 – 760 nanometres – a nanometre is one millionth of a millimetre. Violet light has the shortest wavelength and red the longest. Infrared radiation has a longer wavelength than red light, but a shorter wavelength than microwaves. Nearinfrared light is often used in fibre-optic communication and night-vision devices. Commercial near-infrared spectroscopy systems for in-line sorting of fruit were first adopted by Japanese packhouses about 30 years ago. Near-infrared spectroscopy relies on light with a wavelength of 750 – 2 500 nanometres. It can measure certain quality parameters such as sugar content, as well as detect some internal defects. Ultraviolet radiation has a shorter wavelength than violet light – around 10 – 400 nanometres. Presumably ultraviolet radiation needs no introduction. See our articles on sunburn in the previous issue of this magazine, should you need a refresher. X-rays have even shorter wavelengths – 10 nanometres to 10 picometres – than ultraviolet radiation. There are 1 000 picometres in a nanometre! Their incredibly short wavelengths enable X-rays to effectively penetrate non-transparent objects such as the bodies and luggage of humans. X-ray scanners are therefore ubiquitous in hospitals and airports. They also have many other uses – a notable example is the X-ray diffraction technique that was key to discovering the double helical structure of DNA.
WHAT IS X-RAY CT?
CT is short for computed tomography. A CT scan is the same thing as a CAT – short for computed axial tomography – scan. Tomography is a method for building up a three-dimensional image from a consecutive series of two-dimensional images. Tomo- is derived from tome, which is ancient Greek for slice or section. X-ray CT obtains two-dimensional images of an object by directing an X-ray beam through it, to a detector. In human medicine, CT scanners rotate the source and detector around a stationary person. A computer assembles the individual images to reconstruct a representation of the body. Read page 84 for results of a recent Hortgro-funded project that applied X-ray CT to study internal browning and water core in Fuji.