Release date: 2016-11-29
IBM's recent achievements in the medical field are gratifying and moving. Following the successful diagnosis of skin cancer with photos, the IBM Research Institute recently released the latest results, using deep learning and neural networks to make great progress in identifying mitosis of cancerous cells.
When diagnosing cancerous cells, doctors mainly analyze the way patient tissue samples are analyzed by biopsy. However, even if these tissues are sometimes as tiny as needles, pathologists need to detect signs of tumor cell disappearance, and also observe the important features of cancerous cells to help doctors to prescribe the right medicine.
When the pathologist analyzes the sample, some typical tissue samples are color-coded with the reagent solution. The results show that the depth of the reagent color and its distribution in the cell tissue can distinguish the type of disease and the degree of disease deterioration.
Breast Cancer Cell Training Samples for the 2016 Cancer Diffusion Assessment Challenge
The pathologist then studied the labeled tissue sample under a microscope. However, this stage is time consuming and the workload is huge. The researchers have to deal with hundreds of samples every day, and such long time and high load work will inevitably lead to a decrease in the correct rate of diagnosis.
With the development of modern medical imaging technology and deep learning, pathologists are in urgent need of computer technology assistance, and computer scientists are making unremitting efforts. In order to verify the application of artificial intelligence technology in the medical field, scientists organized a hackathon challenge.
A few weeks ago, with the support of the Utrecht University Medical Center, Eindhoven University of Technology, Beth Israel Deaconess Medical Center and Harvard Medical School, the organizers held the "Tumor Diffusion Assessment Challenge" in Athens, Greece. (Tumor Proliferation Assessment Challenge, TPAC), as a chapter event of the 2016 MICCAI International Conference.
159 teams from all over the world downloaded the 500 breast cancer cell images provided by the medical school on the first day of the event. As a training sample, the data set exceeds the resolution of 50000*50000 pixels. It is true that this challenge was a fierce battle until the end of the game, and only 14 teams submitted the results.
One of the teams came from the IBM Swiss Lab and the IBM Brazil Lab. This international team of Tibetans, Crocodile and Crocodile is composed of French, Hungarian and Greek, and participated in the “Adaptive Algorithm-Based Mitosis Detection Challenge†challenge. The competition lasted for several months and it took a whole summer, but the rewards were rewarded. They won the second place in this competition, only 0.004 points from the first place.
IBM researchers Erwan Zerhouni, Maria Gabrani and David Lanyi use deep learning and neural networks to solve problems in cancer
"It is extremely difficult to manually identify the mitosis of cells. If so, give it to the computer to solve it," David Lanyi said. Before working at IBM, he worked in the field of deep learning at the Zurich Institute of Technology.
“In July of this year, we began to train the characteristics of tissue samples through a neural network-based deep learning algorithm. The main task of training is to find the nuances of negative and positive tissue samples. After a period of training, machine learning The effect is remarkable."
Erwan Zerhouni said, “This was almost an impossible task five years ago. Currently, it takes an hour for the algorithm to diagnose a 5600*5600 image, which we can continually optimize in subsequent studies. Compress time costs to less than 20 seconds while diagnosing any type of cancer."
“We have managed to combine MICCAI's latest deep learning techniques to meet in-depth analysis of omics data (including genomics and proteomics) to provide patients with more accurate medical diagnosis.†From IBM Brazil Labs and participated in this The challenge's Matheus Viana is often thinking about the future of this project. At the same time, the team is preparing to share the findings of breast cancer imaging analysis with researchers at the IBM Haifa Laboratory.
Cancer is just one type of disease that IBM is investigating in the field of medical imaging. Dr. Tanveer Syeda-Mahmood, an IBM member and medical imaging specialist, learned about the team's progress plan and plans to collaborate with it to introduce deep learning methods into medical screening research. This is of great benefit to the study of radiology and cardiology. Similarly, pharmacological and pathological experts often encounter similar challenges in the study of visual fatigue. The research results of Syeda-Mahmood will be presented at the annual meeting of the Radiological Society of North America next week.
Source: Lei Feng Net
Weighing Scale,Weighing Inidcator,Weighing Load Cell,Weighing Equipment,Electronic Scale,Weighing Controller
Changzhou Satidi Import and Export Co., Ltd. , https://www.guanjiejt.com