Chapter 562
Chapter 568
Dr. Jiang’s words were serious and realistic.
The coverage of this problem is too wide, definitely not only Lu Yu’s words that can be solved by removing metal artifacts, and even in a sense, Lu Yu’s words just now are like …. What should I do if I get sick? Then cure the disease.
If you want to say that there is something wrong with this statement, it is not a big problem.
But the feasibility is too poor.
Persuasion is also clearly insufficient.
If you want to make a wave of large-scale blood exchange with domestic nuclear magnetic resonance device technology, you have to come up with some real materials to really convince those people! Otherwise, if something goes wrong, who will be responsible for it?
Heard Dr. Jiang’s questioning voice.
Lu Yu did not show the slightest panic.
He just spoke up again:
“Removing metal artifacts is not so much a means as our goal, and my method is more inclined to deal with the special types of artifacts produced during CT imaging of objects containing metal substances.”
“You definitely understand this, and you may even know more than me, so I won’t introduce this more.”
Lu Yu smiled slightly and said, ”
Metal artifacts are mainly caused by the loss or distortion of projection data caused by the high absorption and scattering of X-rays by metal substances.”
“The methods for removing metal artifacts can be divided into projection domain methods, image domain methods and deep learning methods.”
“The projection domain method is to perform operations such as interpolation, filtering, or iterative reconstruction on the projected data to compensate for or repair missing or distorted data.”
“The image domain approach becomes a reconstruction image for denoising, segmentation and substitution to eliminate or mitigate artifacts.”
“And the most important thing is the deep learning method!”
“It uses neural network models to learn the mapping relationship from artifact-free images to achieve automatic and efficient metal artifact removal!”
“Neural networks?” Dr. Lin next to him, and even the researcher, frowned.
An inexplicable look flashed in their eyes.
Of course, they have all heard of the term neural network. But most of the time, neural network technology is generally used in artificial intelligence level technology!
It can even be said that basically the vast majority of fields involving artificial intelligence technology cannot escape the scope of neural network technology.
This researcher from the Zhurong 2 Aerospace Research Institute, although he does not know artificial intelligence algorithm technology, but neural networks and deep learning are two words that are thunderous…
But he had heard it countless times.
So when he heard Lu Yu mention this word, he couldn’t help but be a little surprised.
“Isn’t that talking about NMR technology? How to use neural networks? Doctor
Jiang frowned and asked,
“Do you want to combine NMR technology with artificial intelligence?” Then, through the neural network, the artificial intelligence deep learning is used to strengthen the judgment of metal artifacts, and then automatically remove which metals are artifacts when the MRI scan is completed?
“If I’m not mistaken, do you mean that?”
His eyes were glued to Lu Yu, waiting for an answer.
“Hmm.”
Lu Yu nodded gently: “Nowadays, our domestic artificial intelligence technology has developed a lot.
“Many similar technologies can really come in handy for devices in some practical fields, so since artificial intelligence technology can be used in industrialized fields, why can’t our medical device technology be combined with artificial intelligence technology?”
This time, he did not wait for Dr. Jiang or Dr. Lin to refute, but said all his views in one breath.
“Of course, if you want to apply artificial intelligence technology to our MRI medical devices, it is not a matter of one day or two, which requires a lot of data accumulation and correction.”
“Only by screening out the high-quality and suitable data can we really let the neural network we build truly achieve the purpose of deep learning, so as to remove metal artifacts in concrete practice.”
“Otherwise, once the neural network builds a learning system that is not intelligent enough, it will greatly increase the possibility of medical accidents.”
“So getting the right data is the top priority we need to do right now, which is why I’m talking to you and your student, Dr. Lin, about this today?”
Old Doctor Jiang’s eyes moved.
He guessed Lu Yu and what he wanted to say next.
“You… Do you want us to call on all hospitals to share data with each other, and build a large enough database for neural networks that can’t be loaded into MRI devices?
He bit his lip, pondering in his mind, not knowing for a moment whether he should approve or disagree.
The application of artificial intelligence technology to such a field is obviously something that Dr. Jiang has never thought about before, he is older, but he knows a lot about young people’s things, and knows that the topic of artificial intelligence is a very hot concept nowadays.
It’s even like Lu Yu said before.
Many traditional industrial fields have begun to carry the practical application of artificial intelligence to increase the efficiency of daily work.
Of course.
The use of artificial intelligence technology mentioned here is not like those science fiction movies as said, the use of some intelligence and intelligence is not significantly different from humans, or even far greater than human intelligent robots to replace human work.
Although this type of so-called intelligent robot has appeared many times in many science fiction movies and science fiction novels, it has even triggered human discussions and research on the crisis of artificial intelligence.
But these things are in reality…. It is far from possible.
That type of artificial intelligence is usually the so-called strong artificial intelligence, and it only exists in the imagination at present.
The vast majority of those used in the actual field are the so-called weak artificial intelligence, intelligent management and application of industrialized modules, intelligent assembly line management… These are also examples of the practical application of artificial intelligence in the industrial field.
It’s just that the current artificial intelligence is still very dependent on data…
Only through the large-scale learning of data can we really make these machines have a lost spirituality, so as to achieve the work that satisfies humans.
This is also why when Lu Yu proposed the words deep learning and neural network, Dr. Jiang guessed what Lu Yu wanted to say almost immediately….
If you really want to combine intelligent artificial intelligence technology with nuclear magnetic resonance devices.
Then the first step is to sort out a large amount of NMR case data to form a database for the corresponding neural network for in-depth learning.
This is the same as human learning.
Only after reading a large number of textbooks and actual cases can it be possible to make your own judgment on the case based on experience.
And if you really want to let the machine carry out the operation of removing metal artifacts, then the requirements for this amount of data will definitely become very large, even the largest hospitals in Kyoto, I am afraid that there is no way to meet this data volume requirements.
Therefore, Old Doctor Jiang guessed what Lu Yu wanted to say.
Since the amount of data in one hospital is not enough to complete the requirements, call on other hospitals to contribute their own data together.
Anyway, if the results can really be made, it will not only benefit the hospitals of one or two, but the patients of the entire Dragon Kingdom!
“You kind of understand what I mean.”
“But what we need is not just the amount of data, just a large enough database, I am afraid it is not enough to meet the deep learning requirements of neural networks.”
Lu Yu said.
“What do you mean?”
Doctor Jiang asked, “In addition to the need to build a large enough database, are there any additional requirements?”
“Teacher, what he said should be the requirements for data quality.” Dr. Lin said suddenly.
“That’s right!”
“That’s what I mean.” Lu Yu turned his head to look at Dr. Lin, and there was a look of satisfaction in his eyes:
“It seems that Dr. Lin, you also have some understanding of artificial intelligence technology?”
“Where, where, what do I know? I’ve just seen similar popular science videos before.
Dr. Lin smiled very modestly:
“After all, this is the age.”
“What’s not more about one+? What is the concept of Internet + to play the Internet, artificial intelligence also has a so-called artificial intelligence + concept.
“Whether these concepts can be put into practice is unknown, but the concept itself can still help us understand some related technologies.”
Although in terms of medical skills, he has a big gap compared to his own teachers.
But… Perhaps it is also due to Dr. Lin’s younger age, his knowledge and grasp of artificial technology is richer than that of Dr. Jiang.
So when Lu Yu mentioned the problem of data, his first reaction was not the amount of data, but the quality of the data.
For the database that artificial intelligence needs to learn, the amount of data alone is extremely large… That’s not enough, you also need to manage the quality of the data!
If the quality is not high enough, then these quantities become so-called toxic data.
Feeding these reading data to a deep artificial intelligence neural network to learn, not only can not have the effect that everyone wants them to have, but may have side effects!
It is indeed a little difficult to understand this point in terms of the concept of artificial intelligence.
But with other examples from life, it’s much simpler. It’s like playing chess, if you play with a smelly chess basket every day, I’m afraid that the technique will not improve, but it will become more and more smelly.
But if you play against different masters every day, carefully and extensively learn different chess scores.
Then it is difficult for you to improve your chess skills!
This is the difference caused by the quality of the data, the same size of the packet, if the quality is different, feed to the same artificial intelligence algorithm model, the final result will be very different!
“So we still need to do a round of sifting through the data?” Old Doctor Jiang figured it out a little.
“That’s right.”
“If the database can be built smoothly.”
“So before throwing the data into the database, we also need to conduct a preliminary screening of the data to screen out the toxic data.”
“Make sure we feed the model…. is a really valid data!
Lu Yu said very seriously.
“If this is to be implemented, will it be difficult?” Doctor Jiang asked again.
He is a person in the medical field, and in theory, Lu Yu should have asked him.
But the current problem in the field of artificial intelligence is that Lu Yu knows more than him, in the end, how toxic data can be counted as toxic data, and how good data can be considered high-quality data?
And how much data is needed to build a suitable and complete database to feed to the artificial intelligence algorithm model?
The answers to these questions are clearly in Lu Yu’s hands.
Therefore, for Old Doctor Jiang, who was eager to figure out all this, he naturally could only throw this doubt to Lu Yu.
“Don’t worry, since I raised this issue to you today, it proves that this problem can be solved.”
“The problem of the reproductive database is precisely an advantage of our dragon country.”
“Advantage? How to say this. Old Doctor Jiang had already faintly guessed what Lu Yu was going to say next, but he still wanted to listen to Lu Yu finish speaking.
“This is the same as the level of our Longwei surgeon.”
“Compared to the vast majority of countries overseas, our dragon country has a larger population, and while maintaining such a high population, it also has medical conditions that are at least above average.”
“This leads to the fact that we have a lot of data and the vast majority of the data quality is passable.”
“It’s not the same as the case with the writers.”
Thinking of this, Lu Yu couldn’t help but smile.
The Dragon Kingdom is large and has a wide population.
In this problem, population resources are also a special advantage, more patients, more cases, more data that can be used as artificial intelligence deep learning!
It’s not like some linguistic AI models from past lives.
In the past life, some of the biggest problems encountered by some language artificial intelligence models in China….
It’s not just that the model of the artificial intelligence model itself is not intelligent enough, but the data fed into it has a large amount of toxic data, which also benefits from a large number of marketing numbers on the Internet.
This problem does not exist, at least in the medical community.
Marketing numbers dare to mess around in terms of news, play headline parties and water injection data… That’s because if you mess around, no one will really come and catch them.
But the medical level is different!
Human life is at stake!
When it comes to human life, who dares to come nonsense?