The AI ​​tool increases transparency in X -rays

New artificial intelligence system ITPCTRL-AI promises to significantly improve the X-rays of the chest By offering both interpretation and control-relating to a long-lasting challenge of AI transparency in medical imaging. Developed by scientists from the University of Arkansas in cooperation with MD Anderson Cancer Center, Models, ITCTTRL-AI, models of pattern visual patterns of radiologists To ensure a decision -making process in accordance with human knowledge.

Diagnostic tools based on AI showed extraordinary accuracy in detecting medical irregularities, such as fluid accumulation in the lungs, enlarged hearts and early symptoms of cancer. However, many of these AI models act as “black boxes”, making it difficult for doctors to understand how the conclusions are drawn.

According to NGAN LE, an assistant professor of computer science and computer engineering at the University of Arkansas, transparency is of key importance for the adoption of artificial intelligence in medicine. “When people understand the process of reasoning and limiting AI decisions, they trust and accept technology more often,” said Le.

EtcTrl-Ai, an abbreviation for interpretative and controlled artificial intelligence, was designed to fill this gap by restoring the way radiologists analyze x-rays of the chest. Unlike conventional AI systems, which simply provide for diagnosis, etcTrL-AI, it generates thermal maps-visual representations of areas on which radiologists focus during their examination. These thermal maps provide a clear view on the AI ​​decision -making process, increasing both trust and interpretation.

To develop this AI model, scientists followed the eye movements of radiologists, browsing the chest X -rays. They recorded not only where the experts looked, but also on how long they focused on specific areas before achieving the diagnosis. The collected data was then used for training, ITPCTL-AI, enabling it to generate heat maps that emphasize key diagnostic regions in the image.

Using these observations based on the look, the AI ​​system filters irrelevant areas before making a diagnostic forecast, ensuring that it will only recognize significant information-like a human radiologist. This decision-based decision-based approach makes, ITPCTL-AI, is much more interpreted than traditional AI models.

To support the development of ITPCTRL-AI, scientists have created diagnosed news ++, the first of its kind data set that adapts medical results with the data of radiologists. Unlike existing data sets, diagnosed news ++ provides detailed anatomical attention maps, setting a new standard for diagnostic transparency based on AI.

Using the semi -automatic approach, the research team filtered and structured data of radiologists, ensuring that each heat map accurately corresponds to medical irregularities. This set of data not only improves the interpretation of artificial intelligence, but also paves the way of future progress in AI medical imaging.

EtcTrL-AI is not the only system based on AI, which develops the transparency of medical imaging. In QDATA we also use grad-cam (mapping of weighted gradient activation) to generate heat maps for mammogram analysis.

At the root of Grad-Cam, he emphasizes the most influential regions of the image that contribute to the decision of the AI ​​model, enabling radiologists to indicate areas of interest with greater precision. This technique ensures that the detection of breast cancer supported by AI remains explained and adapted to medical knowledge. Thanks to integrating visual explanations based on the heat map, both the ITCTRL-AI and QDATA solutions, AI, increase trust and usability in clinical conditions.

Transparency in the diagnosis supported by AI is not only technical progress-it is an ethical necessity. The ability to clarify AI's decision is crucial to ensure honesty, alleviate bias and maintain responsibility in healthcare. Thanks to legal and ethical problems related to AI in medicine, etccTL-AI, it offers a model that allows doctors to take responsibility for the diagnosis supported by AI.

The research team is now working on strengthening ITPCTL-AI to analyze three-dimensional CT scans, which require even more complex decision-making processes. Taking into account information about the depth and wider anatomical structures, the AI ​​system can further improve diagnostic precision in critical medical applications.

To encourage further research and adoption, the source code, models and a set of data with annotations will be made available publicly. This initiative aims to determine a new reference point for transparency and responsibility based on AI in medical imaging.