As the world's number one player Ke Jie lost to Google's artificial intelligence system AlphaGo at 0:3 on May 27, 2017 in China's Wuzhen Goqi Summit, artificial intelligence (arTIficial intelligence) once again caused people's attention. Artificial intelligence refers to the intelligence exhibited by systems that are artificially manufactured. The research of artificial intelligence is highly technical and professional, and the branches are very deep and incompatible, so the scope is very wide. In the medical field, artificial intelligence has also played an important role, especially in the field of tumor diagnosis and anti-cancer drug research and development. What is the application of artificial intelligence in the field of diagnosis and treatment of breast cancer? What new developments are there? As a breast surgeon, the author has made a classification of the application of artificial intelligence in the field of breast cancer in recent years.
I. Artificial intelligence and imaging diagnosis of breast cancer
Mammography, ultrasound, MRI and other imaging techniques have become an important means of breast cancer detection, staging, efficacy evaluation and follow-up. In the 1980s, computer-aided diagnosis (CAD) technology developed rapidly in medical imaging diagnosis with advances in computer technology, mathematical algorithms, and statistics.
CAD refers to the quantitative analysis and processing of lesion characteristics through imaging, medical image processing technology and other means to assist the imaging physician to discover and analyze the lesions, avoiding the limitations of subjective factors such as clinician experience and knowledge level. The errors brought by sex, thus improving the accuracy and efficiency of diagnosis The workflow of CAD system is generally divided into several processes: data preprocessing-image segmentation-feature extraction, selection and classification-recognition-result output.
According to the similarity of algorithm function and form, machine learning generally includes support vector machine (SVM), fuzzy logic, artificial neural network (ANN), and K nearest neighbor algorithm (K-nearest neighborsalgorithm). Different types of algorithms, such as random forests, have different advantages and limitations.
In 1998, the ImageChecker mammography CAD system developed by American R2 Company was approved by the US FDA and became the first CAD system to be put into clinical application. Breast CAD is currently widely used in the screening of breast cancer by X-ray photography. The related research mainly focuses on improving the accuracy of calcification and mass detection. Among them, X-ray photography has a higher detection rate of microcalcification. The detection rate of the mass is affected by the density of the gland.
Italian scholar Parmigegani et al. developed an artificial neural network-expert system (ANN-ES) to improve the identification of mammography results, which can help radiologists obtain higher rates of breast cancer diagnosis. In 2016, Stephen T. Wong and Jenny C.Chang, the Houston Methodist Hospital in the United States, developed a natural language processing (NLP) software algorithm that accurately acquired the key features of mammography in 543 breast cancer patients. And associated with breast cancer subtypes, the diagnosis rate is 30 times that of the average physician and the accuracy rate is as high as 99%.
The world's first breast ultrasound CAD equipment B-CAD was successfully developed by Medipattern of Canada, and was approved by the FDA in 2005 to be sold in the United States. Its standardized breast examination results grading reporting system increases the accuracy of ultrasound diagnosis and can assist physicians. Improve the rate of diagnosis of breast cancer. Chabi et al. reported that CAD for breast ultrasound is highly sensitive and is a useful tool for primary imaging physicians to improve the diagnostic level of breast malignancies, but there are still problems with lower specificity. In recent years, high-tech medical device companies in China have also been involved in intelligent breast ultrasound systems and developed corresponding products, and their application value has yet to be tested in the market.
Similarly, the CAD system can assist in the visual assessment of breast MRI and provide useful additional information. In 2003, Comfirma introduced the first commercial breast MRI-CAD software, CADstream, which analyzes the pharmacokinetic parameters of the contrast-enhancing contrast agent based on MRI-enhanced scanning, and detects the lesions in combination with morphological parameters. And qualitative, the software is still widely used in MRI screening for breast cancer. B ttcher et al believe that the CAD system has high specificity (100%) for MRI assessment of invasive breast cancer response to neoadjuvant chemotherapy, but the lower sensitivity of 52.4% is not a substitute for visual imaging assessment. Song et al found that the CAD system has obvious advantages for MRI assessment of multifocality of invasive breast cancer, but it is not effective for assessing the metastatic state of lymph nodes.
Second, artificial intelligence and pathological diagnosis of breast cancer
1. Artificial intelligence and pathological diagnosis of lymph node metastasis of breast cancer
Conventional breast cancer pathological diagnosis is performed after the tissue has been treated by fixation, dehydration, waxing, embedding, etc., and the tissue section is made. After staining, the pathologist analyzes the lesion characteristics to determine the diagnosis result. Pathological diagnosis is also known as the "gold standard" for diagnosis. The evaluation of sentinel lymph node in breast cancer is extremely important for TNM staging of breast cancer patients and clinical treatment such as axillary lymph node dissection. Manual pathological examination of sentinel lymph nodes is time-consuming and laborious, and the metastases are small. It is even more difficult to get the right diagnosis. The artificial intelligence involved in the pathological diagnosis of breast cancer lymph nodes is generally the final reading section. Similar to the CAD-assisted breast cancer imaging diagnosis mentioned above, artificial intelligence intelligently processes pathological images through specific algorithms, and through training and optimization of algorithms, to develop a high-precision and high-efficiency pathological recognition algorithm model. aims.
In March 2017, scientists from Google Brain, Google Inc. and Verily Life Science used the artificial intelligence technology of the conoluTIonal neural network to perform mammography on 130 pathological sections. Detection of cancer lymph node metastasis. Before the formal test, the scientists provided a number of pathological sections of tumor tissue and normal tissue in advance, and divided the slice images into tens of thousands to hundreds of thousands of small areas of 128 × 128 pixels for artificial intelligence learning. At the same time, a human pathologist took the same test for 30 hours. As a result, artificial intelligence achieved an accuracy of 88.5% and the accuracy of pathologists was only 73.3%.
Google is not the only technology company that is committed to applying artificial intelligence technology to breast cancer pathology image analysis to improve diagnostic efficiency. In June 2016, at the International Biomedical Imaging Symposium, a research team from the BethIsrael Deaconess Medical Center (BIDMC) and Harvard Medical School developed a deep learning-based artificial intelligence technology that will The combination of pathologist's analysis and artificial intelligence automatic calculation and diagnosis method improved the accuracy of breast cancer sentinel lymph node metastasis to 99.5%. Dr. Andrew Beck then set up a diagnostic technology company called Path Artificial Intelligence to develop and apply artificial intelligence technology to help pathologists diagnose diagnoses faster and more accurately. In March 2017, the Dutch multinational electronics company Philips announced its cooperation with the company. DeepCare is a technology company in China that uses artificial intelligence and deep learning technology for the identification and screening of medical images. According to reports in the literature, in 2016, the artificial intelligence algorithm developed by it has a sensitivity of 92.5% for the pathological section diagnosis of lymph node metastasis of breast cancer.
2. Artificial intelligence technology and pathological diagnosis of fine needle aspiration cytology (FNAC)
FNAC is one of the important early diagnostic methods for breast cancer. Artificial intelligence technology can also play a role in improving the diagnostic accuracy of fine needle aspiration (FNA). Fiuzy et al. reported the development of a new algorithm based on the integration of evolutionary algorithm (EA), genetic algorithm (geneticalgorithm, GA), fuzzy C-means (FCM) and artificial intelligence technologies such as ANN. A diagnostic accuracy of up to 96.58% was achieved for 205 breast cancer FNA test samples. With the help of Neurointelligence software, Subbaiah et al. established the ANN model, which accurately identified FNAC samples from all 52 cases of breast fibroadenomas and 60 cases of invasive ductal carcinoma of the breast.
From the above results, artificial intelligence is far better than humans in the pathological diagnosis of breast cancer, especially lymph node metastasis. Does that mean that artificial intelligence can completely replace human experts in imaging and pathology? According to Martin Stumpe, Google's technical director, artificial intelligence still lacks rich knowledge and experience compared to human pathologists. False positive misjudgments can occur, and abnormal classifications such as inflammation that have not been trained can be detected. For the time being, in order to obtain the best clinical diagnosis and improve the consistency of pathologist diagnosis, it may be a feasible method to integrate these artificial intelligence technologies into the clinical work as an auxiliary tool for pathologists. Although there is still a distance from the laboratory to the clinic, the author believes that the process of combining the two may be achieved quickly.
Third, artificial intelligence assisted anti-breast cancer drug development
Founded in 2008, American biopharmaceutical company Berg Health has developed an artificial intelligence platform for rapid screening of patient tissue samples including pancreatic cancer, bladder cancer and brain cancer, and analysis of the gap between the corresponding genomic information and biomolecular metabolic pathways. To find potential drug targets. This method of using data as a starting point and using data to generate a series of putative targets and corresponding drugs is exactly the opposite of the pattern of routine drug development involving a large number of trial and error processes, which is expected to significantly reduce the cost of new drug development. In October 2016, the US Department of Defense announced the recruitment of the company to analyze healthy and diseased tissue models with trillions of data points by analyzing 13,600 tissue samples from 8,000 breast cancer patients, using artificial intelligence techniques to analyze these Models of molecular features in the model to identify unknown breast cancer subtypes and to develop more targeted new breast cancer drugs.
Unlike the previous case, understanding the mechanism by which tumors develop resistance may be another way to develop new anti-tumor drugs. In November 2016, IBM partnered with the Massachusetts Institute of Technology and Harvard University's Broad Institute to launch a $50 million, five-year cancer genome project to acquire and analyze approximately 10,000 cancer patients. Tumor genomic data to help humans better understand the molecular mechanisms of cancer resistance, predict which drugs may be resistant to which drugs, and aim to develop a new generation of anticancer drugs that can overcome drug resistance.
4. Medical aids based on the super artificial intelligence computing platform
The aforementioned Watson artificial intelligence platform from IBM combines a number of technological innovations including information analysis, natural language processing and machine learning. IBM applies its technology potential to business, in the medical field and commemorates Sloan-Kate. The Lynn Cancer Center has jointly developed Watson for Oncology (WFO) to help physicians provide better personalized cancer treatment options for patients. At the San Antonio Breast Cancer Conference in the United States on December 9, 2016, Somashekhar et al reported that the WFO was found in a double-blind comparison of WFO and the multidisciplinary team at Manipal Hospital in India for treatment recommendations for 638 patients with different stages of breast cancer. The standard recommendation treatment and reference for the two opinions and the Indian oncology multidisciplinary expert team reached 73%, but in terms of time-consuming, human experts on average takes 20min, and WFO from the extraction of analytical data to give treatment recommendations, on average only It takes 40s. According to the information provided by IBM, WFO can read 3469 medical monographs, 248,000 papers, 69 treatment plans, 61540 experimental data, 106,000 clinical reports within 17 seconds, and finally propose according to the patient index information input by the doctor. A preferred personalized treatment regimen. So far, WFO has taken extensive medical information including NCCN's clinical guidelines, more than 300 medical journals, and more than 200 textbooks.
Currently, WFO has served and served nearly 10,000 patients in seven countries around the world (China, the United States, South Korea, Thailand, Singapore, India, the Netherlands). As early as August 2016, WFO entered the Chinese medical field. IBM and 21 hospitals in Beijing, Shanghai, Guangzhou, Zhejiang, Fujian, Yunnan and other provinces reached the cooperation intention of Watson tumor solution. At present, some hospitals have already The trial operation of the "Waston Joint Consultation Center" was started to provide doctors with the latest treatment trends, help doctors choose the best treatment plan and help train young doctors. The main target patients are breast cancer and lung cancer patients.
5. Other applications of artificial intelligence
In 2012, Korean scholars reported a predictive model based on SVM to predict breast cancer recurrence in breast cancer patients in South Korea within 5 years after surgery. In 2015, French scholars developed a method based on fuzzy logic selection, which used artificial intelligence to screen the genetic characteristics of breast cancer and successfully applied it to the pathological grade of breast cancer to judge the prognosis of patients. American biosensor company Cyrcadia Health has developed a personalized wearable smart underwear iTbra that can detect early breast cancer using machine learning algorithms and artificial intelligence recognition programs by detecting small temperature changes in the breast tissue. The product is still in beta. The above information is a general analysis of the current global application of artificial intelligence in various fields related to breast cancer. the author thinks:
(1) From the current clinical application, the application of artificial intelligence is better in the fields of medical imaging and pathology. The work content of these two fields happens to be an important part in the diagnosis of breast cancer. I believe that in the near future, Artificial intelligence will play a more powerful role in these two areas.
(2) Artificial intelligence assisted clinicians provide diagnosis and treatment programs for patients, which may have certain positive effects on improving the accuracy and efficiency of diagnosis and improving the competitiveness of hospitals. However, clinical medical behaviors have extremely distinctive human characteristics, and artificial intelligence cannot replace them. Physicians communicate and comfort patients, respecting patient privacy and protecting patient privacy is also a potential problem. Moreover, the basic issues such as the pricing of artificial intelligence medical expenses, the relationship between artificial intelligence and patient medical insurance, how medical data is legally open, and the legal responsibility for artificial intelligence in medical disputes are still subject to official guidance. .
(3) At present, artificial intelligence is still in the stage of “weak artificial intelligenceâ€. The author believes that there is still a long way to go before the goal of fully integrating into the daily work of clinicians, large-scale use on a global scale, and effectively narrowing the gap in medical standards between different regions. Despite this, medicine has always been a technology-driven field, and the development of future technologies will follow the path of artificial intelligence-assisted-small-substituting-mostly-subverted or “liberated†physicians. Just as AlphaGo can win all the masters of Go, it is a bold idea. In the "strong artificial intelligence" stage, the medical robot with all-round, fully automatic and full-disciplinary integration will also surpass the general practitioners. Just like in the movie Prometheus, a fully automatic surgical robot that can automatically perform laparotomy on patients will definitely be invented, just a matter of time.
(4) At present, the artificial intelligence industry is experiencing explosive growth in the medical field, while China is still in its infancy in the field of artificial intelligence, and there are few original key technological innovations in the medical field. The Chinese government is highly concerned about the development of artificial intelligence. Li Keqiang, Premier of the State Council, first mentioned "artificial intelligence" in the "Government Work Report" made in 2017, emphasizing the need to accelerate the development and transformation of artificial intelligence technology in China. Artificial intelligence technologies and capitals based on China's massive patient population should be actively invested in the medical field for the benefit of Chinese patients and physicians. For breast surgeons, it is unlikely that a supercomputer doctor will suddenly save a breast cancer patient, but I believe that more and more breast cancer patients and physicians will be able to make indirect or direct improvements from artificial intelligence. Benefit from it.
Glass Metal Sealed Cap
Glass Metal Sealed Cap
YANGZHOU POSITIONING TECH CO., LTD. , https://www.pst-thyristor.com