There are two answers to this question:
People who are optimistic about AI will say YES, because it will take a few years for AI to replace those mediocre doctors, and will replace those who are Below average, but will not replace those above average. ) doctor.
A conservative doctor who is not optimistic about AI will say NO, TA is really from the heart, naively that AI will not replace any doctor. In fact, the most need for AI in the future is the Doctor of Below average, and the least understood and most unacceptable person.
Recently, AI is very hot, not only in the investment world, but also in the academic world.
A good AI company can easily get the money, and even the AI ​​company has got the money. It is no wonder that some clinicians say that some AI projects are purely money-saving and burning money, and certainly cannot enter the clinic in the future.
However, the good AI project is quite reliable. Recently, Nature has published several articles in which AI has won various academicians.
Look at these eye-catching article titles and you will understand.
"Humans are defeated... Diagnosing breast cancer, 30-hour pathological analysis is not as accurate as Google AI"
Recently, scientists from Google, Google Brain and Verily have developed an artificial intelligence that can be used to diagnose breast cancer. It has outperformed professional pathologists.
Insiders know that the accuracy of pathological diagnosis depends heavily on the level of the pathologist. Even for the same patient, the diagnosis given by different pathologists tends to be very different: a 2015 paper found that the difference is different. The consensus rate of pathologists for breast cancer diagnosis was only 75.3%. In some atypical breast cancers, the consensus rate of diagnosis has dropped to 48%, less than half.
Have you been afraid of this situation? And in China we still lack a lot of pathologists.
The pathologist must undergo years or even ten years of training to gain sufficient experience and become a qualified pathologist. It is even more difficult to become an excellent pathologist. In areas where medical resources are insufficient, I think To get a diagnosis is a luxury.
In order to solve the bottleneck of pathological diagnosis, scientists at Google and Verily made an attempt. They divided the image of a single pathological slice into tens of thousands to hundreds of thousands of small areas of 128x128 pixels, each of which may contain several tumor cells. Subsequently, they provided pathological sections of many tumor tissues and normal tissues for artificial intelligence learning. In the end, this artificial intelligence mastered a pixel-level technique that identifies the pixels labeled as “tumors†in a single small area, marking the entire small area as a “tumor areaâ€, which effectively maps the tumor tissue. Separate from healthy organizations.
After learning, this artificial intelligence ushered in actual combat. The scientists invited a pathologist and asked him to play a game with artificial intelligence. The pathologist spent a full 30 hours carefully analyzing 130 slices and giving his diagnosis. The accuracy of this pathologist was 73.3% in the subsequent ratings based on sensitivity (how many correct tumors were found) and false positives (how much normal tissue was diagnosed as tumors). The answer sent by artificial intelligence is 88.5%, which is better than humans.
“The FDA first approved a cardiac NMR imaging AI analysis softwareâ€
On January 10, 2017, according to the FDA official website, it first approved Cardio DL, a software for cardiac magnetic resonance imaging AI analysis. This software will be used for medical image analysis and for traditional cardiac MRI scan image data. Provides automated ventricular segmentation analysis that is as accurate as traditionally required by radiologists to perform manually.
This deep learning-based artificial intelligence medical image analysis system has performed data validation of thousands of cardiac cases, and the results produced by the algorithm are comparable to those of experienced clinicians.
It is reported that this artificial intelligence cardiac MRI medical image analysis system has not only obtained the approval of FDA510(k), but also obtained CE certification and approval in Europe, which indicates that the software will be allowed for clinical use.
Nature (Hazlett et al. 2017) is heavy: AI beats doctors in the early diagnosis of childhood autism!
Recently, under the leadership of Heather Hazlett, a psychiatrist at the University of North Carolina (UNC) Chapel Hill, artificial intelligence has taken another step in the field of disease diagnosis. The deep learning algorithm they developed predicts whether autistic high-risk children (with an autistic brother or sister) before the age of 2 will be diagnosed with autism after 2 years of age, with an accuracy of 88%. The accuracy is only 50% of the traditional behavioral questionnaire (Charman 2014).
Artificial intelligence once again defeats humans in the field of disease diagnosis.
"AI robot, after learning 2186 lung cancer maps, winning the pathologist"
On August 16, 2016, Nature News published a study by researchers at Stanford University School of Medicine: Computers can be trained to be more accurate than pathologists in assessing lung cancer tissue sections.
The researchers used 2186 images of lung cancer gene maps from patients with adenocarcinoma and squamous cell carcinoma. The database also contains information on the grade, duration, and survival time of each patient after diagnosis.
The researchers then used these images to train computer software programs to determine more cancer-specific features that could not be observed by the naked eye—nearly 10,000 personality traits vs. evaluation features commonly used by hundreds of pathologists. These features include not only the size and shape of the tumor cells, but also the shape and texture of the nucleus and the spatial relationship with adjacent tumor cells.
Dr. Snyder, a professor of genetics at Stanford, said: "In the aftermath, everything is justified. Computers can evaluate even small differences in thousands of samples more accurately and quickly than humans."
"AI re-enrolls the cover of Nature: Diagnosing skin cancer, accuracy is comparable to experts"
Researchers at Stanford University used deep convolutional neural networks to develop pattern-recognized AI through extensive training, enabling computers to analyze images and diagnose disease.
The training computer database consists of 129,450 skin lesion images and corresponding textual descriptions covering 2032 skin diseases. The “reference answers†​​for diagnosis are provided by dermatologists who rely on non-invasive image analysis and tissue biopsy.
After that, the computer ushered in the "graduation exam." The researchers provided a number of images of skin lesions that had not appeared in the training data set to the trained computer and 21 practicing physicians. These images were confirmed by tissue biopsy. The result of the diagnostic competition is that the computer is as accurate as a human doctor and sometimes better than a human doctor.
How does AI artificial intelligence replace doctors?
Artificial intelligence is not a panacea, but it does exceed human capabilities in certain disciplines and fields, replacing doctors' work and even completely replacing doctors.
If the data or images used to diagnose the disease or judge the prognosis can be standardized, quantified, and structured, it can basically be done with artificial intelligence. After the algorithm is established, the machine can be continuously learned and accumulated, and gradually improved. Will eventually defeat humanity.
From the current application, the fields of artificial intelligence application are dermatology, pathology and imaging.
The dermatology department in Taiwan and some European and American countries are the most preferred departments for medical students when they graduate. Because they work relatively easily, they have a good income and they don't have to work night shifts.
Unfortunately, artificial intelligence is coming, and it is likely to replace the work of many people. Over time, dermatologists will replace many people's work.
At present, the most lack of doctors in China is the pathology department. Unfortunately, over time, the work of pathologists and imaging doctors may be taken away by artificial intelligence, and the level of AI will be higher than most general practitioners.
Looking at the obstetrics and gynaecology department I am working on, there are a lot of routine work that can be done with artificial intelligence in the future. There is no need for so many doctors.
Screening for cervical cancer: After the doctor collects cervical exfoliated cells, the machine can automatically produce tablets to automatically determine whether there are cancer cells. In terms of benign and malignant cervical cells, at least 80-90% of the pathologist's work can be replaced by AI.
Obstetric ultrasound: In the screening and diagnosis of fetal malformation, the general practice in North America is to be retained by the general Sonographer ultrasound technician in accordance with the standard cut-off screenshots, and then the MFM maternal medical expert review report. In theory, the reading of these ultrasonic cross-sections can be done by AI. The principle is similar to the interpretation and judgment of AI on CT and MRI films.
Fetal heart monitoring: Interpretation and judgment of fetal heart monitoring results can also be completed by AI.
The machine is more reliable than humans, the machine is more precise than humans, and the machine will not fatigue. With the continuous advancement of algorithms and the continuous accumulation of data, the level of artificial intelligence will become higher and higher, and will evolve from the current help for human judgment to the replacement of humans. Make judgments.
This trend is irreversible and irresistible, and the FDA cannot stop it. There will be a Breaking Point tipping point in the future. After this tipping point, there will be a cliff-like fall in the regular workload of the doctor.
The future scenario will be: Above average doctors are AI assistants, and Below average doctors are AI assistants.
Of course, those non-standardized, full of uncertainty, and manual operation of clinical work are still irreplaceable by AI.
Tetanus Shot,Tetanus Vaccine,Hepatitis B Injection,Hep B Vaccine
FOSHAN PHARMA CO., LTD. , https://www.foshanpharma.com