Click "i Dark Horse" above and select "Pinned Public Account"
Dark Horse Think Tank, a must-read for entrepreneurship
i Dark Horse News (Xiaomi) July 6 news, hosted by the entrepreneurial dark horse "2018 China Unicorn Summit" held in Beijing today, it is estimated that technology CEO Chen Kuan attended the event and delivered a keynote speech.
Chen Kuan said in his speech that although AI medical treatment can produce great value in reducing the rate of missed diagnosis and misdiagnosis, graded diagnosis and treatment, medical insurance control costs, health economy, etc., there is currently a lack of unified industry standards. He believes that clinical-grade AI products must be "robust", "easy to use" and "safe".
The following is an excerpt from a speech edited by i Dark Horse:
First of all, I would like to thank the organizers for their invitation, and it is a great honor to give you a brief introduction to the landing and actual use of AI in the medical field. ”
In fact, you have seen that artificial intelligence has blossomed in various fields, including security, new retail, autonomous driving, and the medical and health field where we speculate about technology. It's solving different problems in different industries, and I think the development of AI in the medical field has a very far-reaching impact on the future of mankind.
1 Worldwide medical resources are insufficient
The first problem facing the landing of AI in all industries is not the problem of technology itself, but what kind of problem it can solve in this field, here, we need to review the current situation of the medical industry. The medical problem does not exist only in China, it is actually a worldwide problem.
The first is the lack of high-quality medical resources. In China, many hospitals are overcrowded, although this is not the case in foreign hospitals, but foreign medical services are very expensive, and many patients do not get good diagnosis and treatment services. There's a reason for that.
For example, to see a CT image of a lung, a doctor may take a long time to read it. The difference between normal lungs and early lesions of the lungs is very subtle, ordinary people to see there is no way to distinguish it, even if it is a doctor with many years of experience in diagnosis and treatment, it is likely that there will be missed diagnosis and misdiagnosis due to fatigue.
For this problem, artificial intelligence is a good solution. In addition, there are many application scenarios of artificial intelligence in the medical industry, such as the medical imaging and auxiliary diagnosis links we are in.
In 2015, when Speculative Technology was just established, there were actually only two companies in the world that were really doing medical AI, medical deep learning and medical imaging. By 2018, there have been a lot of medical AI companies mushrooming, and there are more and more explorers in academia and industry.
I remember when we first entered the field of medical deep learning in 2015, I searched for the word "deep learning" on Google, and I only found one article, which was written by a Japanese professor at my alma mater, the University of Chicago, called "Virtual Neural Networks and Medical Imaging", and he did not set himself in the field of deep learning at that time. Now, you can also search for it, basically you can see tens of thousands of articles, including academia and industry. From this point of view, the field of medical deep learning has achieved an explosive development.
2 Lack of uniform industry standards
But any outlet industry, including medical AI, will have some problems.
For example, many so-called expert systems at present, fish-eyed beads, say that they are deep learning. For example, a large number of plagiarism and imitation, just like the earliest lung nodule products made by speculative technology, later there were a large number of imitations. There are also data security issues, and patient privacy is not well protected. These problems, in the final analysis, are because there is no unified industry standard, and no one plans what kind of product is a good AI product.
The development of medical AI to the present, in the speculation of technology and the efforts of all enterprises in the industry, artificial intelligence has actually been integrated into the daily diagnosis and treatment path of doctors.
I think good AI products must be repeatedly polished by clinical trials and continuously iterated according to the needs of doctors. It requires quality teachers, good data sources, and good data standards, high-quality deep learning models. In this process, repeated practice is required to repair.
I think that the criteria for judging AI products are mainly reflected in the following three aspects:
1, robustness. A good AI product must be able to maintain relatively high stability and accuracy in different medical fields and medical environments, so robustness is extremely important.
2. Ease of use. The value of AI lies in helping humans release more capacity, and if the use of AI products is very troublesome and complex, it is useless.
3. Security. The use of AI products must be under the condition of ensuring the safety of the entire medical system.
So, I think robustness, safety, and ease of use are necessary for a clinical-grade AI product.
3 Industries need to be rooted in landing development
So far, several of our industry lines have begun to land in the industry. We came across a case when one of our AI products went live.
After the doctor saw the lung images, the doctor's instructions were: there was no obvious abnormality in both lungs. But our product circles an abnormal point when identifying lung images, which doctors say may be false yang. Later, the patient was recommended for follow-up examination and found that it was indeed cancerous. AI does have a great help for doctors in identifying lesions.
In addition to lung products, we are also doing corresponding product research and development in the breast, bone, heart, and liver. So far, The Products of Speculative Technology have completed nearly 13,000 cases of auxiliary screening of lung cancer every day, and AI is already one of the largest diagnostic systems in the world. The products of Speculative Technology have landed in nearly 150 top three hospitals around the world, including the United States, Japan, Europe and so on.
China is currently implementing graded diagnosis, but the requirements for grass-roots hospitals are: grass-roots hospitals must have the ability to find early lesions. But the contradiction is that the lower the probability of early lesions being missed, the higher the need for diagnostic capabilities. Grassroots hospitals are often the hospitals with the worst medical equipment and medical resources, and AI medical treatment can bring great value to society, reduce the probability of missed diagnosis of early lesions, diagnose diseases in advance, win time for patients, and improve the cure rate. We hope that artificial intelligence can help sink some high-quality medical resources and medical diagnosis to the grassroots level.
I think artificial intelligence can eventually produce different values in different industries, in the medical industry, artificial intelligence in graded diagnosis and treatment, medical insurance cost control, health economy and other aspects can produce great value, which is also the direction that the entire industry needs to work on in the future.
So far, Speculative Technology has gradually become a full-scenario, all-type, all-type medical institution service platform in the medical industry. We also hope that artificial intelligence technology can eventually bring high-quality medical resources to thousands of households.
* Let entrepreneurship no longer be lonely, improve the success rate of ordinary entrepreneurs, welcome to pay attention to i dark horse.