Four-word long text restore IBM Watson: Interview with many doctors, AI experts, related companies

Section 1: Stars suddenly falling into a vortex

Throughout the history of global business, it is not difficult to find that successful companies have been prosperous for a long time, even for a hundred years, because they can continue to launch milestone products and services, and pull out the growth curve again and again.

For Blue Giant IBM, there are countless such products, such as System/360, laser myopia surgery, RAM, PC and the familiar Thinkpad.

At present, Watson is undoubtedly a product that has been given historical responsibility. Through its cognitive platform "Watson", IBM hopes to realize its transformation to "cognitive computing" through the power of artificial intelligence.

According to John Kelly, senior vice president of IBM's cognitive solutions and IBM research, IBM Watson Health's focus is on applying next-generation AI technology to cancer treatment. In the direction of betting cancer, IBM Watson Health has developed three unique cancer treatment solutions to assist doctors around the world in treating patients:

Watson product line

Before 2017, Watson was still a popular star project, and the "breakup" with the Anderson Center caused it to be "pushed down the altar". Negative news followed:

February 2017, operating M. D. The University of Texas at the Anderson Center announced the closure of a partnership with IBM to pay IBM $39 million in compensation for the initial $2.4 million contract on the contract.

In May 2017, Chamath Palihapitiya, a senior technology investor and founder of venture capital firm Social Capital, even directly bombarded CNBC in May: “Watson is a joke.”

. . . .

In May 2018, IBM Watson Health was exposed to layoffs;

In July 2018, IBM Watson Health was exposed by the media to recommend 'unsafe and incorrect' cancer treatments;

In October 2018, Deborah DiSanzo, head of IBM Watson Health, announced his departure.

Organizing the media views, you will find that its doubts about Watson products can actually be attributed to the following four categories:

Fosun, the chief artificial intelligence expert of Fosun Pharma, and CTO Deng Dao, who explained the product of Waston, explained to the arterial network that Watson is facing the current "difficulties" for four reasons:

One is to over-promote Watson to replace the doctor, to surpass the doctor, to surpass the doctor's cognition, and to issue treatment plans for intractable diseases. Such publicity has rapidly raised the expectations of IBM Watson. In the industry, there is still a lack of uniform testing standards, and the final clinical effects of the products have yet to be evaluated. Excessively exaggerated market propaganda is not good for the long-term healthy development of products;

Second, Watson's current theoretical system is still not perfect enough to take on the function of machine reading;

Third, in the process of R&D technology, the number of people involved in product development and research is insufficient, and the number of real medical records is small;

Fourth, IBM's internal strategic planning was improper. In August 2015, IBM spent $1 billion to acquire medical imaging company Merge Healthcare. It did not have outstanding achievements in the field of imaging, and the path of resource integration was unclear.

According to public information, Watson can support the following aspects, including but not limited to:

·Understanding natural language

·Big data understanding and analysis

· Dynamic analysis of various assumptions and problems

·Fine personalized analysis capabilities

·Optimize questions and answers based on relevant data

· Refining insights and discovering new operating modes in a short period of time

· Learn in iterations and explore optimized solutions

According to the understanding, Watson's processing logic is an application of open question and answer technology such as natural language processing, information retrieval, knowledge representation, automatic reasoning, machine learning, etc., based on hypothetical cognition and large-scale evidence collection, analysis and evaluation. DeepQA.

In a long article translated by Professor James Hendler, Lei Feng revealed the process of Watson's realization based on "associated knowledge." In short, after the doctor enters information about the patient's medical condition, the application recommends the treatment by analyzing published studies that may be relevant. The IBM Watson for Oncology process includes analyzing patient medical records, providing treatment options, and sequencing:

1. Analyze patient medical records, including structured and unstructured data;

2. Provide treatment plan options, by analyzing a variety of medical data, IBM Watson for Oncology offers several treatment options for each patient, the doctor may choose in these programs;

3 . Sort the program, sort the treatment options, and indicate their medical evidence.

Watson's diagnostic process, arterial network mapping

Watson “learns” by constantly adjusting its internal processing procedures to get high-resolution, correct answers on certain issues, such as radiographic images that reveal cancer. The correct answer must be known so that the system can be informed, when to do what is right, and when to do something wrong. The more training problems the system can handle, the higher the hit rate.

Through analysis, provision of programs and optimal program sequencing, the final patient gets a handful of reports on the recommendations of a rich cancer treatment program. Among them, there are several recommended schemes, consideration schemes and non-recommended schemes. Behind each recommendation, Watson “doctors” will indicate the source and basis and arrange them in order of credibility for the treatment of a physician. If the doctor chooses a treatment plan, it will also give information about the survival rate, adverse reaction rate, and drug interactions used to help the doctor evaluate the efficacy and risk of the program.

In fact, most of the criticisms against Watson refer to statements that exaggerate publicity and are too optimistic about Watson's prospects.

If Watson has not made significant achievements so far, one of the most obvious obstacles is that it requires certain types of data to be "trained", which are usually either very scarce or difficult to access. This is not a unique problem for Watson, which is a common problem in the entire medical machine learning field.

Although the lack of data has affected Watson's development speed, it has a greater impact on IBM's competitors. In the algorithm and model training of medical AI, the best way to obtain data is to work closely with large medical institutions, but these institutions are often very technically conservative.

Due to the rigor of medical care, AI products require a large amount of clinical data validation to be recognized. Although major hospitals are now open to medical AI, the competition between similar products is extremely fierce. It is often the case that there are many similar products for doctors to use in a department, and it is difficult to obtain clinical data.

The biggest feature of Watson's "doctor" is that it can learn and progress quickly. In 2017, Watson “doctors” added 4 new cancers and 6 new treatments, and all indicators were continuously upgraded and improved. By 2018, Watson's "doctor" new treatment program has covered breast cancer, lung cancer, rectal cancer, colon cancer, stomach cancer, cervical cancer, ovarian cancer, prostate cancer, bladder cancer, liver cancer, thyroid, esophageal cancer and endometrium. 13 cancers of cancer. According to reports, Watson "doctors" study papers have included papers from Hong Kong experts.

On the one hand, it is a question of IBM and Watson. On the other side, IBM also responded positively to the negative public opinion that has been hit.

Just three days after the Wall Street Journal reported on the Watson dilemma, John Kelly, IBM Cognitive Solutions and Senior Vice President of IBM Research, quickly responded:

"IBM has a lot to be proud of, including groundbreaking research on Watson Health. Unfortunately, some media reports, including a report published by The Wall Street Journal on August 11, have distorted and ignored some facts, suggesting IBM has not yet made 'sufficient' progress in applying the benefits of artificial intelligence to the health care arena. It is imminent to clarify the truth."

According to the Wall Street Journal, the largest AI product in the Watson Health portfolio is Watson for Oncology, which typically costs $200 to $1,000 per patient, and in some cases requires consulting fees.

Since 2012, the New York Memorial Sloan Kettering Cancer Center has been helping IBM train the software (not using the software for patient care). The hospital's experts collaborated with IBM engineers to rank relevant features of medical history such as tumor location and coexistence conditions, and to rank medical studies for specific therapies. Then evaluate Watson's ability to match the test case to the treatment and help the engineer adjust the output until it is consistent with the doctor's judgment.

According to reports from IBM, Watson is working closely with leading cancer research organizations such as Memorial Sloan Kettering and the Mayo Clinic to develop and improve cognitive solutions. They are currently used in 230 hospitals and healthcare facilities worldwide. By the end of June 2018, the number of patients had reached 84,000, almost double the number of patients who received the service by the end of 2017. ”

In the face of real-life case questions, John Kelly presented a series of Watson applications and report data:

A report by a doctor at the Mayo Medical Center at the annual meeting of the ASCO American Society of Clinical Oncology said that after the Watson Clinical Trial Matching Solution was implemented, the percentage of enrolled in breast cancer trials increased by 80% (up to 6.3 per month). The patient was 3.5 in the previous 18 months).

Dr. Thaddeus Beck and the Highland Oncology Group report that the Watson Clinical Trial Matching Solution reduced clinical trial matching time by 78%.

Dr. Somashekhar and Manipal Hospital said in an "Annals of Oncology" earlier this year that the consensus rate of the breast cancer treatment program for Watson's oncology solution and the treatment plan proposed by the hospital's multidisciplinary oncology committee reached 93%; they recently said They have applied Watson's oncology solutions to all complex cases in the Multidisciplinary Oncology Committee, changing the recommended treatment regimen for 9%-11% of patients.

Dr. Michael Kelley and the Department of Veterans Affairs have just renewed our contract with Watson Gene Solutions. To date, nearly 3,000 veterans with cancer stage 4 have received treatment supported by this solution.

Dr. William Kim and the University of North Carolina Lineberger Cancer Center published a study in which the Watson Gene Solution found new, actionable genetic mutations in 32% of patients.

In IBM's response, John Kelly emphasized that the role of technology is to help doctors provide better care and treatment for patients. The core question IBM has to address is: “Can Watson help oncologists make more effective treatments for their patients?” The core is “help” rather than “alternative”.

This view is somewhat similar when the arterial network talks with some medical artificial intelligence companies. Some insiders say that as a start-up, there will be a company that uses Watson as a learning object to learn the advantages of the product; in addition, Watson Most of the problems now come from media propaganda mistakes. They believe that IBM itself is not intended to diagnose doctors when it launches this product. Artificial intelligence can only act as a doctor's assistant. And for a technology company like IBM, their products should not be too far from the clinical development process.

Although Dr. Deng Wei believes that IBM's products have certain mistakes in propaganda, technology, strategy, etc., he also positively affirmed Watson's "historical achievements" in the application of artificial intelligence in the medical field - changing the methodology of medical treatment.

Taking the CDSS (Clinical Decision Support System) clinical decision support system as an example, the researchers at the University of Pittsburgh began research in this field around 1970. The main methods used by researchers at that time were from medical textbooks and medical literature. The medical rules are extracted and the rules are expressed as the formal logic of "if-then". Enter the patient's symptoms, find the corresponding if, and then infer the disease according to then.

Two years later, in 1972, a professor at Stanford University began a similar study called MYCIN. MYCIN is also mainly the if-then rule base, but later the if-then rule base has a resounding new name called "Expert System".

If-then rules, it is black and white, and it is clearly defined. Later, the probability was introduced into the rules, and the network structure was used to link many rules together. This is the technology of the 1990s, and the Bayesian network, also known as the causal network. The "Bayesian" network is very beautiful in mathematics, but it is very complicated in the actual application process. It can't find a good application to land. So after a while, the Bayesian network is hot.

Until 2011, IBM Watson appeared. In the beginning, IBM Watson was a research project at IBM Research. The research team began researching natural language processing in 2006. They taught the machine and extracted the words "Portuguese, Vasco da Gama, arriving at Calicut on May 20, 1498". From other literatures, the phrase "Kalikat is located in southwestern India" is extracted. Then the two sentences are concatenated and the conclusion is made, "Portuguese, who landed in India in 1498."

The most famous result of this topic is that Watson participated in the American Knowledge Quiz TV competition "Jeopardy" in 2011 and defeated human players. This game is very important. It actually declares the arrival of the AI ​​artificial intelligence application era, which is a major milestone in the history of artificial intelligence. On the road of turning scientific and technological achievements into profitable products, after an evaluation, IBM finally chose AI.

The choice of medical care is the right turn for IBM, and the market for medical services is huge. From the massive medical records, the experience of clinical diagnosis and treatment of human doctors is excavated. This methodology is inductive. The method used to extract rules from the medical literature is the deductive method. IBM Watson changed the methodology used by artificial intelligence.

The history of modern science shows that changes in methodology are likely to bring about earth-shaking changes. IBM Watson refines clinical diagnostic experience from massive medical records, rather than extracting and reasoning medical rules from the medical literature, which is a paradigm shift. IBM Watson is actually leading the cognitive revolution.

At present, many people regard Google's Google Medical Brain project as an industry leader in artificial intelligence medical. The Google Medical Brain project also explores the clinical pathways of human doctors from a vast array of medical records. And in April this year published a paper in Nature magazine, systematically depicting the entire project plan of Google Brain. Although Google Medical Brain is slightly better in terms of detail, it is consistent with the Watson methodology.

Section 3: The real experience of doctors

Shanghai Ten Hospital and Zhoukou City Hospital of Traditional Chinese Medicine introduced Watson in August 2017 and February 2018, respectively, and mainly used in the Oncology Department. The arterial network interviewed Director Xu Qing of the Department of Oncology of Shanghai Tenth Hospital and Director Zhang Yueqiang of the Department of Oncology of Zhoukou City Hospital of Traditional Chinese Medicine to understand the use of Watson in their department.

Xu Qing told the arterial network reporter that since the introduction of Watson in the Oncology Department of Shanghai Tenth Hospital, nearly 650 cases of tumor patients have been assisted, accounting for about 50% of all outpatients, involving cancer types including colorectal cancer, stomach cancer, lung cancer and other multiple cancers. Kind.

Doctors usually recommend Watson when the patient's condition is more complicated, but since Watson's decision-making is not included in the medical insurance, the patient only uses Watson if economic conditions permit. "Some patients are also using Watson because of their admiration." Xu Qing said that the introduction of Watson has improved the appeal of the oncology department of Shanghai Tenth Hospital to some extent.

The proportion of patients with oncology in Zhoukou City Hospital of Traditional Chinese Medicine was lower than that of Watson. Zhang Yueqiang told the arterial network that only about 10% of patients used Watson. Zhang Yueqiang believes that whether patients use Watson is related to the complexity of the disease itself and the economic strength of patients.

In terms of accuracy, both Xu Qing and Zhang Yueqiang stated that Watson's essence is to assist decision-making and treatment tools, and the level of consistency between the treatment recommendations and clinical disease guidelines should be used as a measure of accuracy. Because Watson has studied a large number of clinical disease guidelines and medical literature in data training, its accuracy is quite high. Zhang Yueqiang said that according to his estimation, Watson's accuracy can reach 90%. "Watson's treatment plan is based on the latest research results, and sometimes the treatment plan it gives is even more reasonable." Zhang Yueqiang added.

However, Watson is not exactly in line with the doctor's expectations. Xu Qing and Zhang Yueqiang said that at present, Watson's localization level can not fully meet the clinical needs. The main performance of Watson's incompatibility in domestic hospitals is that the proposed drugs are not listed domestically.

As early as 2017, the arterial network reporter interviewed the first doctors who used Watson in China. Gu Xidong, the chief physician of breast surgery in Zhejiang Provincial Hospital of Traditional Chinese Medicine, said that for doctors, Watson has four uses: (Original: "Interview with China Batch use Watson's doctor, he thinks AI has 4 major uses, 2 points are insufficient.)

The first use is to select the best treatment plan by studying empirical speech;

The second use is to reduce the misdiagnosis of doctors;

The third use is to provide doctors with a new treatment plan as a reference;

The fourth use is to help train young doctors.

For the lack of Watson, Gu Xidong said that the first is that Watson is positioned as a doctor to assist in the adjustment of the patient's real life situation, only to recommend the target pathological indicators, but the situation of cancer treatment is very complicated, and Not the best treatment plan is the one that the patient can accept. In many cases, the doctor needs to adjust according to the actual condition of the patient, and to convince and comfort the patient, which Watson can't do.

The second is that Watson is currently unable to combine Chinese and Western medicine. Gu Xidong said that Chinese medicine has gradually received attention, and some Chinese doctors have some shadows of Chinese medicine practitioners when they are conditioning patients. At present, Watson does not have the ability to do so.

Therefore, learning more domestic clinical cases can make Watson better localized.

Wang Dong, director of the Robot Micro-Invasive Center of the Sichuan Provincial People's Hospital, also said in public that doctors hope to achieve true AI in every aspect of the diagnosis and treatment. The Watson system is relatively mature in the diagnosis.

Prof. Xiaohui Xiao from the Changzheng Hospital affiliated to the Second Military Medical University published a view on AI from the perspective of doctors/users: First, there is no problem of falsification in AI, but it is currently immature. The tumor chemotherapy program only considers the European and American guidelines and some of the expert experience. The guideline guides the treatment's defects as “paper talk”, and needs to integrate more expert experience in the future. Second, in the past few years, AI's ability to diagnose and treat has been despised by “senior experts”. Even if experts use Da Vinci surgical robots, they do not feel the threat of AI. Third, the development of AI must be accompanied by ups and downs. In this process, we need to adapt and adjust. "The era of AI medicine has come to the fore. Whether you like it or dislike it, it is there, and sooner or later it will play a leading role. This is determined by the will of the people, but ultimately it will not be transferred by the will of the people."

In addition, in the interview, Zhang Yueqiang also said that artificial intelligence is a general trend, hospitals and doctors should be open to medical AI products into the department. Now medical AI itself is still in its infancy, and it is believed that medical AI products will be smarter and more practical in the future.

Section 4: Arterial Network Exclusive Interview with Watson China Distributor

In order to obtain the appearance of Watson in the real medical scene in China, and whether the outside question is true. The Arterial Network reporter has repeatedly tried to contact Zhang Wenming, the general manager of the Watson Health Division Greater China and Asia Pacific, and the other party did not respond to the matter. Subsequently, we contacted Watson China's general agent Baiyang Intelligent Technology (hereinafter referred to as Baiyang), and Baiyang Chief Marketing Officer Wang Biquan accepted our interview.

Wang Biquan believes that many questions and accusations against Watson are untrue and inconsistent with the real situation. He emphasized that, first of all, Watson is not a robot, but an artificial intelligence system with the ability to understand, reason, analyze, learn and interact. Second, Watson has no prescription rights. It only provides decision support for oncologists, allowing doctors to save a lot of time to care for more patients. "When Watson is in Holmes, Watson is the doctor's best helper."

"They have never used Watson's tumors, have never conducted on-site research with real users of the hospital, and never investigated the real use scenarios of Watson's tumors with IBM or official operators. They just collected some network information. And speech, it is a misunderstanding of it in a seemingly professional way, we do not recognize this behavior, it may affect Watson to some extent." This is Watson's first public response.

In addition, Wang Biquan responded to the "Watson's wrong medicine" incident that was previously spread on the Internet.

The beginning of the incident was the publication of a confidential document within IBM by the US medical media STAT, which documented the clinician’s strong criticism of Watson and showed that Watson’s medical advice and underlying technology had serious problems.

One case was a 65-year-old man diagnosed with lung cancer with severe bleeding symptoms. In this regard, Watson's medical advice is to receive chemotherapy and use the drug for the treatment of cancer bevacizumab. However, one of the side effects of bevacizumab is that it is prone to bleeding.

Subsequently, MSK responded that the "65-year-old male lung cancer patient" was a fictional case given by a cancer center doctor when training Watson, just to train Watson's decision-making ability. Wang Biquan pointed out that in the real world, bevacizumab is a prescription drug that must be sold, adjusted, and used by a physician. At the same time, when a doctor enters patient information in the Watson system, he or she will be asked to choose whether the patient has hemoptysis. “Yes”, the system will automatically filter drugs that are likely to cause bleeding in patients.

Wang Biquan said that since Watson landed in China, the hospital's demand for it is very obvious. The hospital has the need to train young doctors, the need to learn international advanced cases, the need to confirm the treatment plan, etc. This Watson can satisfy.

At the same time, Chinese cancer treatment also requires Watson to provide standardized, evidence-based reference, to avoid the patient's long journey. In addition, hospitals are very willing to accept because real-world needs fit well with Watson's true product value.

"The difficulty of exceeding expectations is mainly reflected in the national lack of understanding of Watson and the singering of foreign media." Wang Biquan appears to be more helpless. “AI is a brand new field. Everyone has a lot of ideas about AI. Some people have a very high expectation for AI, and some people are a little afraid and rejected. So a new thing is brought to this market. There will definitely be one or another controversy."

Wang Biquan added that although the biggest pain point of medical AI is supported by the national level, the charging items in some cities have not yet been clarified, which is not conducive to the promotion of new technologies and new products.

Regarding the large-scale layoffs of Watson, which had caused widespread concern before, IBM had publicly responded that after a period of large-scale acquisitions and migrations to the IBM Cloud, it was necessary to rationalize the business, so personnel adjustments were made. This is normal in business. IBM also has hundreds of job vacancies and continues to recruit in key areas such as data management, analytics and artificial intelligence.

In March 2017, Watson Health signed a contract with Baiyang, and Watson for Oncology officially entered the Chinese market.

It is understood that Watson was developed by IBM and has been trained by the world's top cancer treatment center to commemorate the Sloan Katelin Cancer Center (MSKCC) since 2011. Currently, it has studied more than 330 medical professional journals, more than 250 medical books, and 27 million articles. Paper research data. In addition to research data, Watson also accumulates data through real-world and clinical cases.

Currently, Watson is the only application-level AI tool and the only AI tool that provides a second treatment opinion. Watson's second treatment opinion is based on standardized treatment advice provided by the MSK decision process.

Watson can help doctors get the latest literature in less time. At the same time, with strong reasoning, analysis and interaction skills, can provide doctors with evidence-based, personalized, prioritized treatment recommendations, and after each recommendation, indicate the source and basis for the treatment of the doctor And it takes no more than 10 seconds to complete this series of work.

As of press time, Watson has reached nearly 80 hospitals in 43 cities in 22 provinces across the country. "According to the hospital and the use of Watson's doctor and patient feedback, Watson's supplementary suggestion function was recognized, and Watson was considered to be quite useful in assisting medical treatment and education for young doctors." Wang Biquan told the arterial network.

According to Wang Biquan, Watson has been deeply involved in clinical work in his hospital, helping the development of mechanics and making clinical decisions more efficient. Specifically, Watson provides evidence-based medical evidence-based treatment advice, participates in MDT discussions, and quickly provides detailed medical decision support materials in a real-life medical environment.

According to reports, Baiyang and IBM are actively exploring with local medical institutions to promote Watson's localization. In the future, Watson will learn more domestic medical guides, literature and real-world cases to better meet the needs of domestic oncologists.

One thing is certain, no matter how the future of Watson's development, human exploration of the medical technology revolution using AI will continue.

Nitrogen cylinder

Nitrogen Cylinder,Green Nitrogen Gas Cylinder,Green Nitrogen Cylinder Bottles,Nitrogen Cylinder Bottles

JIANGSU NEW FIRE FIGHTING TECHNOLOGY CO.,LTD , https://www.ffo2cylinder.com