Open Source, Open Mind: Baidu and LinkingMed Help Doctors in Medical Images

Healthcare Author: Mengyao Zhang Editor: Eddie Turkson Mar 05, 2020 11:41 AM (GMT+8)

A Chinese startup, based on Baidu’s deep-learning open-source platform PaddlePaddle, is providing CT screening service to a Hunan-based hospital, hoping to facilitate the diagnosis process.

LinkingMed's AI-driven diagnosis system. Image credit: Courtesy of LinkingMed

LinkingMed has launched its AI-based detection model for pneumonia CT-screening and prediction based on the Baidu’s PaddlePaddle (Parallel Distributed Deep Learning), EqualOcean learned from the medical imaging firm.

A Hunan-based hospital affiliated to Xiangnan University became the first medical institution to use the system.

This AI-equipped diagnosis system can detect and contour the lesion, picture the diagrams of the density of the two hemisphere lungs, and visualize a series of quantitative benchmarks such as quantity and volume percentage. All these tasks can be completed within less than one minute, with an accuracy rate of  92% - 97%, LinkingMed claims.

So far, LinkingMed has not initiated any commercial plans for this AI-based pneumonia detection system.

“The motivation to launch this AI medical diagnosis model is only to help fight against the spread of coronavirus,” Mr. Ryan Zhang, the CEO of LinkingMed, said. “We will unconditionally contribute our AI power to containing this plague, as much as we can,” he added.

The high-tech medical solution provider focused on solving the compatibility issue with different types of scanning equipment by cooperating with research entities. At the development stage of the algorithm used by the system, the Xiangnan University-affiliated hospital provided professional clinical instructions regarding the data annotation, module design and set strict acceptance criteria to satisfy various medical conditions. This online AI screening system will be deployed at multiple hospitals in Hubei, Chengdu, and other severely-affected areas.

Founded in 2016, LinkingMed is an AI technology enterprise in the field of oncological radiation therapy. The Beijing-based medical high-tech company sells techniques and cloud services related to the organic contour, target contour, and radiotherapy to medical institutions. By leveraging the internet and cloud platforms, it also provides remote collaboration and relevant website services for oncological doctors and physicists.

During the past four years, LinkingMed has received d four funding rounds, including the recent one worth CNY 40 million in October 2019. Linear Venture is the serial investor for LinkingMed from Series Pre-A finance to the last Series A finance.

Medical imaging is a niche business given lightweight and comparably less attention as Chinese medical device makers are promoting their ‘globally competitive’ strategy. Now this area is increasingly highlighted, as the novel coronavirus continues to spread and consumes a lot of medical facilities and doctors across China. The domestic high-tech medical device market rose in 2016, a fierce tussle participated by United Imaging (联影医疗), Huiyihuiying (汇医慧影) and VoxelCloud (体素科技). Most players use essentially indifferent AI technology while base the competency on different sources.

These providers still need to consider some fundamental issues to apply AI in daily life. The first is developing the algorithm for the specific problem, for instance, coronavirus event. The other is high-quality data annotation, which takes time and hard work. Additionally, the lack of industry standards, quality protocols, and practice guidelines stand as the major problems ahead of the industry.

For patients, time is life. CT imaging plays a critical role in identifying the infected patients, serving as a surrogate to PCR (polymerase chain reaction) diagnostic, according to a previous report. In the Huoshenshan (‘Fire God Mountain) hospital, one of the frontline hospitals, the CT lung screening by InferVision (推想科技) helped with the lab’s capacity as the number of suspected cases rose quickly. 

To facilitate the detection process and guarantee the test results, LinkingMed incorporated the AI learning framework into the CT imaging process by utilizing Baidu’s PaddlePaddle. Injected with the AI’s deep-learning power, this open-source online diagnosis model can improve the efficiency and ease the pressure on the clinical doctors. 

Before 2020, ‘Artificial Intelligence’ was a buzz word. Now, the outbreak of coronavirus is pushing many key industries to accelerate their AI processes, especially, in the urgent demand for medical facilities. The epidemic has forced Ali Health (阿里健康), Yitu Technology (依图科技), Huawei (华为), Deep Wise (深睿医疗), and other tech companies to introduce their AI-equipped diagnostics, in different aspects, aligning a joint wholistic force to ease the infection. 

The PaddlePaddle of Baidu is an open-source deep-learning Chinese platform for industrial applications. Similar to Google’s TensorFlow, PaddlePaddle comprises core learning-framework, model bank, development tool kits, and online services supporting over 1.5 million developers in businesses. 

The option to cooperate with PaddlePaddle also indicates LinkingMed’s motivation for its AI diagnostic system. “To curtail the current tension in Korea, Iran, and Japan, we want to provide technical support through Baidu’s platform for international medical institutions and developers,” says Mr. Zhang, “We also will keep an open mind to work with different medical entities, including online healthcare. The goal is to inject technology, whilst fulfilling our social responsibilities. ”