Swarm artificial intelligence learns to detect cancer, lung disease and COVID-19

0
145


Credit: public domain CC0

Communities benefit from sharing knowledge and experience among their members. Following a similar principle — called “swarm learning” —an international research team has trained artificial intelligence algorithms to detect blood cancer, lung disease, and COVID-19 in data stored in a decentralized manner. This approach has advantages over conventional methods, as it inherently provides privacy preservation technologies, which facilitates analysis between scientific data sites. Swarm learning could significantly promote and accelerate collaboration and information exchange in research, especially in the field of medicine. Experts from the DZNE, the University of Bonn, the information technology company Hewlett Packard Enterprise (HPE) and other research institutions report this in the scientific journal Nature.

Science and medicine are increasingly digital. Analyzing the resulting volumes of information, known as ““- It is considered a key to better treatment options.” Medical research data is a treasure. They can play a crucial role in the development of personalized therapies that are tailored to each individual more accurately than conventional treatments, “said Joachim Schultze, director of Systems Medicine at the DZNE and professor at the Institute of Natural Sciences. Life and Medicine (LIMES) at the University of Bonn: “It is essential that science can use this data in the most complete way and from as many sources as possible.”

However, the exchange of medical research data in different locations or even between countries is subject to data protection and data sovereignty rules. In practice, these requirements can generally only be implemented with significant effort. In addition, there are technical barriers: for example, when large amounts of data have to be transferred digitally, data lines can quickly reach their performance limits. In view of these conditions, many they are locally confined and cannot use data available elsewhere.

The data remains in place

In light of this, a research collaboration led by Joachim Schultze tested a new approach to evaluating research data stored in a decentralized manner. The basis of this was the still young “Swarm Learning” technology developed by HPE. In addition to the IT company, numerous research institutions from Greece, the Netherlands, and Germany participated in this study, including members of the “German COVID-19 OMICS Initiative” (DeCOI).

Swarm Learning combines a special type of information exchange between different nodes in a network with “machine learning” toolbox methods, a branch of artificial intelligence (AI). The basic axes of machine learning are algorithms that are trained on the data to detect patterns, and which, in turn, acquire the ability to recognize patterns learned in other data as well. “Swarm learning opens up new opportunities for collaboration in both medical research and business. The key is for all participants to be able to learn from each other without having to share confidential data,” said Drs. Eng Lim Goh, Senior Vice President and Chief Technology Officer for Artificial Intelligence at HPE.

In fact, with Swarm Learning, all research data is kept in place. Only algorithms and parameters, in a sense, lessons learned are shared. “Swarm Learning meets data protection requirements in a natural way,” stressed Joachim Schultze.

Collaborative learning

Unlike “federated learning,” in which data also remains locally, there is no centralized command center, the Bonn scientist explained. “Squad learning is done cooperatively based on rules that all partners have agreed upon in advance. This set of rules is captured in a chain of blocks.” It is a kind of digital protocol that regulates among the partners in a binding manner, it documents all events and all parties have access to them. “The blockchain is the backbone of Swarm Learning,” Schultze said. “All members of the squad have equal rights. There is no central power over what happens and the results. So in a sense, there is no spider that controls the data network.”

Therefore, AI algorithms learn locally, specifically from the data available at each network node. The learning outcomes of each node are collected as parameters across the blockchain and the system processes them intelligently. The result, i.e. the optimized parameters, is transmitted to all parties. This process is repeated several times, gradually improving the ability of algorithms to recognize patterns at each node in the network.

Lung images and molecular characteristics

Researchers are now providing practical evidence of this approach by analyzing X-ray images of the lungs and transcriptomes: the latter are data on the gene activity of cells. In the current study, it focused specifically on immune cells circulating in the blood, that is, white blood cells. “Data on gene activity in blood cells is like a molecular footprint. They have important information about how the body reacts to a disease,” Schultze said. “Transcriptomes are available in large numbers, as are X-ray images, and they are very complex. This is exactly the kind of information you need for artificial intelligence analysis. This data is perfect for testing. Swarm learning “.

The research team addressed a total of four infectious and non-infectious diseases: two variants of blood cancer (acute myeloid leukemia and acute lymphoblastic leukemia), in addition to tuberculosis and COVID-19. The data included a total of more than 16,000 transcriptomes. The swarm learning network on which the data were distributed usually consisted of at least three and up to 32 nodes. Regardless of the transcriptomes, the researchers analyzed about 100,000 chest X-ray images. These were from patients with fluid accumulation in the lung or other pathological findings, as well as from people without abnormalities. This data was distributed across three different nodes.

A high success rate

The analysis of both transcriptomes and X-ray images followed the same principle: first, the researchers fed their algorithms with subsets of the respective data set. This included information on which of the samples came from patients and which from undiscovered individuals. Recognition of learned patterns for “sick” or “healthy” was then used to classify other data, i.e., it was used to classify the data into samples with or without disease. Accuracy, that is, the ability of the algorithms to distinguish between healthy and sick individuals, was about 90 percent on average for transcriptomes (each of the four diseases was evaluated separately); in the case of X-ray data, they ranged from 76 to 86 percent.

“The methodology worked best in leukemia. In this disease, the signature of gene activity is particularly surprising and therefore easier to detect by artificial intelligence. Infectious diseases are more variable. However , the accuracy was also very high for tuberculosis and COVID-19. For X-ray data, the rate was slightly lower, which is due to the low quality of the data or image. ” , commented Schultze on the results. “Our study therefore shows that Swarm Learning can be successfully applied to very different data. In principle, this applies to any type of information for which pattern recognition using artificial intelligence is useful. “Whether it’s genome data, X-ray images, data from brain images or other complex data.”

The study also found that Swarm Learning performed significantly better than when network nodes learned separately. “Each node benefits from the experience of the other nodes, although only data is always available. The concept of Swarm Learning has passed the practical test,” Schultze said.

A vision of the future

“I am convinced that swarm learning can give a big boost to medical research and other data-driven disciplines. The current study was just a test. In the future, we aim to apply this technology to Alzheimer’s and other neurodegenerative diseases, ”Schultze explains. dit. “Swarm Learning has the potential to change the game and can help make the wealth of medical experience around the world more accessible. Not only research institutions, but also hospitals, for example, could come together to form these swarms and therefore share information for mutual benefit “.


Artificial intelligence tracks leukemia


More information:
Warnat-Herresthal et al., Swarm Learning for decentralized and confidential clinical machine learning, Nature (2021), DOI: 10.1038 / s41586-021-03583-3

Citation: Artificial intelligence with swarm learns to detect cancer, lung disease and COVID-19 (2021, May 26) recovered on May 26, 2021 at https://medicalxpress.com/news/2021-05-ai- swarm-intelligence-cancer-lung .html

This document is subject to copyright. Apart from any fair treatment for the purposes of private study or research, no part may be reproduced without written permission. Content is provided for informational purposes only.





Source link