How AI and X-rays To Detect Explosives Could Also Identify Cancers
How AI enhance X-rays to detect concealed explosive and potentially tumors, wall breach by their textures
Researchers at University College London (UCL) recently published a paper in Nature Communication, where they demonstrated how a new X-Rays method married with artificial intelligence is capable of identifying even small amounts of explosives hidden in electronic objects. The authors stated that the algorithm under test conditions achieved 100 percent accuracy. Potentially this technique could also be adapted for other applications such as finding tumors, structural cracks, and so o
The unknown rays
X-rays are high-energy electromagnetic radiation (the wavelength is between 10 picometers to 10 nanometers). Although x-rays were generated by other experiments, it is attributed to Röntgen for the discovery (so much so that in some countries such as Germany, Hungary, Denmark, Sweden, etc., they call them Röntgen rays).
Two months after the discovery, Röntgen published a paper (1995). It was an immediate success: 49 papers and 1044 articles were published in 1996 (with the prestigious journal Science alone devoting 23 articles to the topic in 1996). Röntgen also obtained the first Nobel Prize as a result of his discovery. Moreover, he was the first to realize its medical use, so much so that the first X-ray in history is an image of his wife’s hand (in truth, his wife was less enthusiastic upon seeing the image, saying, “I saw death.”)
X-rays were immediately used in the medical field, with the introduction first of radiography and then in the therapeutic field with radiotherapy. The use of x-rays today goes beyond medicine, so much so that they are now also used for:
- X-ray crystallography: for protein and other biological molecules
- X-ray astronomy
- Industrial radiography
- Airport security and Border control uses to investigate the potential threats in luggage and trucks.
In recent years we have become accustomed to queues for security checks, where our luggage goes through an x-ray machine. Anyone who has taken the plane knows that they have to take computers and other electronic gizmos out of their luggage. In fact, it is difficult to be able to identify explosives that have been hidden inside laptops or other items.
How to find concealed explosives.
UCL researchers hid small amounts of explosives (semtex, C4) inside various objects (hair dryers, mobile phones, and laptops). After that, these items were placed inside luggage with other items (as if they were the trunk of an ordinary traveler). The new technique showed 100 percent accuracy in identifying explosives.
“This is a radically different way of inspecting materials and objects by analysing textures, and allows us a new way of detecting illicit materials” — declaration by study authors (source)
The scientists used a specially constructed machine with masks — sheets of metal with holes drilled into them, which split the beams into an array of smaller beamlets — to scan the bags instead of using ordinary x-ray machines, which attack objects with a uniform field of x-rays. The beamlets were scattered at angles as small as a microradian (roughly one 20,000th the size of a degree) as they moved through the bag and its contents. The scattering was examined by AI, which had been trained to identify the texture of particular materials from a particular pattern of angle changes.
In other words, the researchers noted that microscopic irregularities or changes in the objects are bending the X-rays, and the new technique allows researchers to detect it.
This new technique improves the signal and enables discrimination between different materials. However, this remains analyzing the image is still a laborious process. so the researchers used different models of convolutional neural networks to analyze the image. In fact, they tested different architectures based on the use of pre-trained models (such as GoogleNet and Inception ResNet) and transfer learning.
Oddly enough, the best results were obtained without image augmentation (rotation, flipping, scaling, altering the contrast, and illumination properties); this may be surprising however, for the researchers, this stems from the particular properties of the materials. The materials analyzed are grainy, and image augmentation techniques such as scaling can cause alteration of the average grain size/distance.
In addition, no false negatives were obtained in the second test. As the researchers note this: “ is encouraging as missing explosives is a greater cause for concern than a limited number of false positives”.
The model is still improvable, and the researchers do not expect the results in the laboratory to be reproducible with the same accuracy under normal conditions. In the article, they point out how this model could also be applied in other fields. In fact, the model and technique focused on the texture difference between two neighboring materials. This could be useful for identifying tumors that are still very small and therefore treatable but hard to be identified by a clinician. The authors noted, for example, that the algorithm could be used to identify small breast cancers that are not noticed because they blend in with healthy tissue or the rib cage.
In addition, the authors say this technique could be used to detect damage, cracks, and wrinkles in buildings and other construction works before they are visible to the naked eye. This would allow for a noninvasive technique to identify potential problems earlier.
“This has the potential to be incredibly versatile, game-changing technology. We’re currently negotiating with a number of companies to explore how it could be put to practical use. There’s really no limit to the benefits this technique could deliver.” — Professor Olivo (source)
Conclusions
Airport security requires that luggage be x-rayed for potential threats. This is not always easy, and the main factor is the experience of the operator. Especially when in small quantities and hidden in other objects, explosives and narcotics are difficult to detect. This new technique together with artificial intelligence will make it possible to identify explosives, narcotics, and exotic animal contraband.
Computer vision will revolutionize the medical diagnosis, and they expect several of the developed algorithms to enter the clinic in the coming years. There are several studies that have focused on trying to use AI for the early diagnosis of various cancers. Several algorithms have reached state-of-the-art and compete with experienced physicians.
On the other hand, computer vision can revolutionize several fields (from materials analysis to airport security). In this study, in addition to developing an interesting algorithm, they have also built a new X-ray technique, showing that the combination allows for mind-blowing results.
If you have found it interesting:
You can look for my other articles, you can also subscribe to get notified when I publish articles, and you can also connect or reach me on LinkedIn. Thanks for your support!
Here is the link to my GitHub repository, where I am planning to collect code and many resources related to machine learning, artificial intelligence, and more.
Or feel free to check out some of my other articles on Medium:
Additional resources
- About the new X-Ray technique and AI model: here, here, here
- AI to diagnose cancer: here, here, here, here, here
- about use of AI in airport security: here, here, here, here, here, here
- about convolution neural network (here, here, here), transfer learning (here, here, here, here) and pretrained models (here, here, here)
