AI with swarm intelligence learns to detect cancer, lung diseases and COVID-19
Welcome back everyone to our blog . Are you all eager to know what today’s topic is? Today we are going to discuss a very trending topic . Come let us read this and know about it.
Reference :https://www.guvi.in/course_login
Swarm learning
Communities benefit from sharing knowledge and experience among their members. Following a similar principle — called “swarm learning” — an international research team trained artificial intelligence algorithms to detect blood cancer, lung disease, and COVID-19 from distributed and stored data. This approach is superior to traditional methods because it essentially provides privacy protection technology and facilitates cross-site analysis of scientific data. Therefore, herd learning can significantly facilitate and accelerate collaboration and information exchange in research, especially in the medical field. Experts from DZNE, the University of Bonn, the information technology company Hewlett Packard Enterprise (HPE), and other research institutes report this in scientific journals.
Science and medicine are becoming more and more digital. Analysis of the resulting amount of information — “big data“ — It is considered the key to better treatment options.” Medical research data is a treasure. Joachim Schultz, Dean of DZNE’s School of Systems Medicine and Professor of the University Institute for Life Medicine (LIMES), said: Bonn “It is important for science to be able to use such data from as comprehensive and as many sources as possible.”
However, the exchange of medical research data between different locations, or even between countries, is subject to data protection and data sovereignty regulations. In practice, these requirements can usually only be implemented with considerable effort. In addition, there are technical barriers. For example, if you need to transfer large amounts of data digitally, your data lines can quickly reach their performance limits. Considering these conditions, many Medical research It is restricted locally and data available elsewhere is not available.
Data Remains on Site
In light of this, a collaborative study led by Joachim Schultze tested a new approach to assessing distributed and stored research data. The basis for this was the still young “Swarm Learning” technology developed by HPE. In addition to IT companies, the study was attended by a number of research institutes in Greece, the Netherlands and Germany, including members of the “Germany COVID-19 OMICS Initiative” (DeCOI).
Swarm Learning combines a special kind of information exchange between different nodes of the network with the methods of the “Machine Learning” toolbox, a branch of artificial intelligence (AI). The essence of machine learning is an algorithm that trains data and detects patterns in it. As a result, you gain the ability to recognize learned patterns in other data as well. “Swarm Learning opens up new opportunities for collaboration in medical research and business. The important thing is that all participants can learn from each other without sharing sensitive data,” said Dr. Senior Vice President and Chief. Eng Lim Goh says. HPE Artificial Intelligence Technology Officer.
In fact, with Swarm Learning, all survey data remains in the field. Only algorithms and parameters are shared — in a sense, lessons learned. “Swarm Learning meets data protection requirements in a natural way,” emphasized Joachim Schultze.
Collaborative learning
Unlike “associative learning,” where the data remains local, there is no centralized command center, Bonn scientists explained. “Swarm learning is collaborative based on rules that all partners have agreed in advance. This set of rules is incorporated into the blockchain.” This is a kind of digital protocol that regulates. Information exchange Document all events between partners in a binding way and have access to them for all parties. “Blockchain is the backbone of Swarm Learning,” says Schultze. “Every member of the flock has equal rights. There is no Central Powers of what happens and what happens. So, in a sense, there is no spider to control the data web.”
Therefore, AI algorithms learn locally, that is, based on the data available at each network node. The learning results of each node are collected as parameters via the blockchain and processed smartly by the system. The result, the optimized parameter, is passed to all parties. This process is repeated multiple times, gradually increasing the ability of the algorithm to recognize patterns at each node of the network.
Lung images and molecular features
Researchers are now providing practical evidence of this approach through analysis of lung and transcriptome x-ray images. The latter is data on the genetic activity of cells. Current research has focused specifically on the immune cells that circulate in the blood, the white blood cells. “Data on the genetic activity of blood cells is like fingerprints of molecules, which hold important information about how an organism reacts to disease,” Schulze said. “Transcriptomes are as bulky and complex as X-ray images. This is exactly the kind of information you need for artificial intelligence analysis. Such data is great for testing swarm learning.”
The research team worked on a total of four infectious and non-infectious diseases. Two variants of hematological malignancies (acute myeloid leukemia and acute lymphoblastic leukemia), tuberculosis and COVID-19. The data contained a total of over 16,000 transcriptomes. The herd learning network to which the data was distributed typically consisted of at least 3 to up to 32 nodes. Apart from the transcriptome, the researchers analyzed about 100,000 chest x-rays. These were from patients with fluid accumulation in the lungs, patients with other pathological findings, and individuals without abnormalities. These data were distributed across three different nodes.
High success rate
Analysis of both transcriptomes and x-ray images followed the same principles. First, the researchers supplied a subset of each dataset to the algorithm. It contained information about which samples were from patients and which were from individuals with no findings. The data was then further categorized using the learned “disease” or “health” pattern recognition. In other words, the data was used to classify the data into samples with or without illness. The accuracy, or ability of the algorithm to distinguish between healthy and sick, averaged about 90% in the transcriptome (each of the four illnesses was evaluated separately). For X-ray data, it was in the 76–86 percent range.
“This methodology was most effective in leukemia. In this disease, the genetic activity characteristics are particularly striking and artificial intelligence is the easiest to detect. Infections are more diverse. Nevertheless, The accuracy of tuberculosis and COVID-19 was also very high. For X-ray data, the rate was slightly lower due to the poorer data or image quality, “Schultze commented on the results. “Thus, our research proves that Swarm Learning can be successfully applied to very different data. In principle, it applies to all types of information for which artificial intelligence pattern recognition is useful. Data, X-ray images, data, etc. From brain imaging and other complex data. “
The study also found that Swarm Learning gave far better results than if the nodes in the network were trained individually. “Only local data is available, but each node benefits from the experience of the other nodes, so the Swarm Learning concept has passed real-world testing,” Schultze said.
