20 March 2025

Specialists from the Warsaw University of Technology, in collaboration with the Polish Lung Cancer Group, have developed an artificial intelligence model based on the world’s largest database of chest X-ray images. This model will assist doctors in diagnosing diseases within the chest area.

“The system is designed to support the doctor in the most tedious tasks, giving them more time to analyze diagnostically important features,” says Przemysław Biecek, the project leader and head of the MI2.AI research team, which focuses on data processing and machine learning.

As the expert explained in a statement to PAP, by searching for similar images, the system — developed as part of the Xlungs project — can quickly scan thousands of reference CT scans, analyzing hundreds of images in each scan to precisely identify pathological changes and important anatomical features.

“The anatomical features measured by the system can be integrated into other diagnostic processes. Just as a blood test is fundamental in diagnosing numerous diseases, so precise, fast, and inexpensive measurement of changes in the chest area could be a breakthrough in screening,” says the leader of the MI2.AI team. He adds that this tool can be integrated with systems already used in treatment, as it is compatible with commonly adopted medical documentation standards.

The system uses a vast number of lung CT images — as many as 40,000 — developed by the MI².AI research team from the Warsaw University of Technology, in collaboration with the Polish Lung Cancer Group. These images were sourced from CD-ROMs containing CT scans of Polish patients from 2010 to 2018, including those from lung cancer screening programs. As a result, an artificial intelligence model based on the largest such database in the world (containing 40 terabytes of data) has been created. This model is designed to support doctors in faster and more effective disease diagnosis. There could be many more similar resources in Poland.

Imagine credit: Freepik

Every year, several hundred million laboratory tests are performed in Poland, with over 60 million of them being imaging tests, such as CT scans. According to the Collective Minds Radiology report, a single CT scan generates between 200 MB and 1 GB of data. A medium-sized hospital generates anywhere from several dozen terabytes (1 TB = 1024 GB) to several petabytes (1 PB = 1024 TB) of data annually, in the form of imaging scans, lab results, and medical records.

In Poland, electronic medical documentation (EMD) has been under development for over a decade — an integrated system that collects patients’ health data. Since July 1, 2021, every doctor or clinic is required to report medical events in this system. However, many medical facilities were already collecting such data independently before that time.

“We often encounter situations where the treatment of a patient is complete, but their test results remain stored in the hospital or clinic database, metaphorically ‘gathering dust’ on shelves,” explains Marcin Luckner, head of the work within the Xlungs project. “However, even if the data collected for a specific case is no longer relevant, comparing it with results from other individuals with the same condition may help doctors spot patterns and trends in the progression of the disease, ultimately improving future treatment. Analyzing hundreds or thousands of test results is a very tedious and time-consuming task, but AI algorithms can assist in this,” he adds.

According to scientists from the Warsaw University of Technology, Poland adds several thousand IT graduates each year, and half a billion e-prescriptions are issued annually. They believe this gives Poland a significant opportunity to become a leader in the development of AI-supported medical technologies. “Polish data allows for better support of local diagnostics than data obtained from places like China. At the same time, its size offers the potential to create solutions at a global level,” state the specialists from the Warsaw University of Technology.

Source:
Nauka w Polsce