With the aggravation of the social population aging, elderly neurodegenerative diseases such as Parkinson's disease have gradually evolved into important social problems. At present, the number of patients with Parkinson's disease in China is about 3 million. The prevalence rate of people over 55 years old is about 1%, and the prevalence rate of people over 65 years old is 1.7%, but among the 1.7% of the population, only 3.75% of patients were aware of their illness at the onset of the disease, and the rate of misdiagnosis of Parkinson's disease was as high as 23.5%.
Why is Parkinson's misdiagnosis rate so high? This is related to the early symptoms of Parkinson's disease. The early symptoms of Parkinson's disease such as trembling, unstable walking, slow and rigid movements, and lack of words are easily mistaken for the normal phenomenon of the elderly, which leads to the patient visit rate. Very low, often missed. Some studies have found that most patients take an average of 10 months from symptom onset to clinical diagnosis. Even in first-tier cities such as Shanghai and Beijing, the treatment rate is less than 40%.
The treatment status of Parkinson's disease is in a state in which the overall treatment rate is not timely, the delay of diagnosis is high, and the treatment rate is low. It is also an important problem facing medicine. If found early, patients can be treated to relieve symptoms, prevent and reduce long-term motor complications, and lead to normal life. However, if the disease is late, the patient suffers from the disease and the quality of life is seriously degraded.
Because the symptoms of early Parkinson's disease are not obvious, the traditional diagnosis requires a series of judgments. Among them, there is a key link based on the completion status of a series of specified actions. One of the traditional diagnostic methods is to evaluate patients one by one through the Parkinson Rating Scale UPDRS.
The doctor instructs the patient to complete the action based on the Parkinson Rating Scale, and then scores item by item based on the patient's completion, which takes about 30 minutes or more. During this period, you may suffer mental damage from self-doubt and doctor's speech due to factors such as physical coordination and tension. This method of diagnosis relies on language communication, and the time cost is high, and the score mainly depends on the naked eye of the doctor, such as the distance, amplitude, frequency of the action, lack of quantitative indicators, and there may be deviations due to subjectivity.
Tencent Medical Artificial Intelligence Lab recently launched a new technology for AI-assisted diagnosis of Parkinson's disease. With the help of this new technology, daily assessment and early screening of motor function of Parkinson's disease can be achieved. The doctor can complete it within 3 minutes. Diagnostic process, diagnostic speed increased by 10 times.
This new AI-assisted diagnostic technology, called the Parkinson's disease motor function intelligent assessment system, based on motion-free video analysis technology without wearable sensors, automatically implements the Parkinson Rating Scale UPDRS score for Parkinson's motion video.
(Photo: from Tencent WeChat public number)
Simply put, the user does not need to wear any sensors, just need to shoot through the camera (common smartphone can be satisfied), do some simple actions of the Parkinson Rating Scale, such as stretching the fist, hand rotation, etc., the system can recognize The key nodes of the body part in the motion video, quantitative analysis of the action indicators, complete the diagnosis process.
Some people will suspect that the same action is different for various reasons. How can the machine distinguish these nuances?
The machine can diagnose subtle movements without the three core technologies of the intelligent evaluation system: dynamic feature capture (predicting the position of the whole body joint through attitude convolution), and timing analysis technology (coherent to ensure the coherence of the whole body joint in time dimension by time series convolution) Sexuality, as well as dynamic analysis techniques (using memory networks and human dynamics models to output reliable motion metrics).
These three core technologies capture and analyze actions and identify subtle differences. According to the Tencent Medical AI Lab, the user's movements can be split into hundreds of identifiable key points according to the body joint points, and then the model is used for identification detection.
(Photo: from Tencent WeChat public number)
For example, a hand will set 21 key points. As the hand performs various actions, the trajectory of the key points becomes different, and there are changes in the data of frequency, distance, angle, speed, and the like. These measured motion statistics are used to train the AI ​​system. After deep learning, it can distinguish the subtle differences in the movements of patients with Parkinson's disease.
Then, through the motion video analysis technology, the captured patient motions are analyzed, and the "quantitative" and "fine" evaluations are performed according to the model. With these reviews, doctors can more effectively evaluate patients with Parkinson's disease, grade patients, and develop more targeted treatment options.
What is the effect of this system diagnosis? Professor Wang Jian, deputy director of the Department of Neurology, Huashan Hospital, Fudan University, and the National Research Center for Geriatrics Clinical Medical Research (Huashan) Parkinson, said: "The current pre-experimental data show that the AI ​​score of the Parkinson's disease motor function intelligent assessment system. The results are very close to the results of the expert manual scores, and the results are fully achieved. Subsequent large-scale formal clinical trials are in the process of active preparation."
At present, this AI diagnostic system has not been officially put into use, and then it will learn and progress through more rigorous experiments. In the future, it is necessary to enable patients to use the ordinary smartphone to self-shoot, and to complete the daily evaluation of the sports function of Parkinson's disease in the family scene.
In addition to diagnosing Parkinson's, Tencent Medical AI Lab is also applying its video analysis technology to cerebral palsy patients' preoperative gait analysis, football player's post-injury recovery training motor function status assessment, and daily exercise performance of the elderly. In the auxiliary diagnosis of sexual diseases.
Other major products of Tencent Medical AI Lab also include clinical assistant decision support system, providing clinical assistant decision support for high-risk and misdiagnosed diseases such as stroke and acute coronary syndrome, as well as ECG intelligent analysis software, using AI technology to achieve ECG monitoring results. Automatic interpretation and early warning, etc., to create more AI+ medical application scenarios.
Dr. Fan Wei, head of Tencent Medical AI Lab, said: “The pure video analysis objectively quantifies the mobility disorder and the effectiveness of patients after Parkinson's disease in daily life. It is good for patients, patients' families and doctors. Things. After engaging in AI medical related research work, I feel the profoundness of medicine. It is a very comprehensive science. It has both scientific and ethical content. AI can solve some of the key technical problems, for example, Improve efficiency and bring convenience, but it can't solve all the problems. The road is still very long, we still need to move on."
Tencent is actively exploring and deploying artificial intelligence technology in different medical scenarios. Tencent Medical AI Lab is an artificial intelligence laboratory for the medical field. It adopts the US-China dual-center model and is currently set up in Silicon Valley, Beijing and Shenzhen. Three branches. The main research direction of the laboratory is based on natural language understanding, medical knowledge map, deep learning, medical imaging, Bayesian network, multimodal analysis and other basic technologies to build medical knowledge engine, medical reasoning engine, clinical auxiliary diagnosis engine, and consultation. Intelligent platform such as dialogue engine.
Mackerel
Mackerel,King Mackerel,Jack Mackerel,Saba Mackerel
ZHEJIANG EVERNEW SEAFOOD CO.,LTD , https://www.evernewseafood.com