Deep Learning Reconstruction Combined with Ultra-Low Pitch Scanning: A Synergistic Strategy for Helical Artifact Suppression in Low-Dose Abdominal CT

Medical imaging has long faced a difficult balancing act. To see inside the human body clearly, doctors rely on X-ray scans that pass through tissue to create a picture. The clearer the picture, the easier it is to spot problems like tumors or inflammation. However, getting a sharp image often requires a higher dose of radiation, which carries its own risks, especially for patients who need frequent checkups. Conversely, lowering the radiation dose to keep patients safe often results in grainy, noisy images where fine details blur together. In the abdomen, this problem is compounded by the way the scanner moves. As the machine rotates around the patient, it spirals forward to capture the whole body. If this spiral is too wide or fast, gaps appear in the data, creating streaks and distortions known as helical artifacts. These streaks can hide the delicate walls of the intestines, making it hard to diagnose conditions like Crohn's disease or bowel obstructions. For decades, radiologists have had to choose between a fast, low-dose scan with poor image quality or a slow, high-quality scan that exposes the patient to more radiation.

A team of researchers at Guiqian International General Hospital in China set out to break this trade-off. They tested a new approach that combines two specific technologies to see if they could work together to solve the problem. The first part of their strategy involves slowing down the scanner's spiral movement to an "ultra-low" speed. By moving the scanner more slowly, it captures a much denser amount of data, filling in the gaps that usually cause those distracting streaks. The second part of the strategy uses a modern computer program based on deep learning. Unlike older computer programs that simply smooth out grainy images, this new program has been trained on thousands of real medical images. It learns to recognize the difference between random noise, unwanted artifacts, and the actual anatomy of the body. The researchers wanted to know if using this slow scan speed together with the smart computer program could produce a clear, low-radiation image of the abdomen that was better than what is currently possible.

To find the answer, the researchers enrolled 90 patients who needed abdominal CT scans for medical reasons. They divided these patients into three groups to compare different scanning methods. One group received a standard scan using a conventional speed and a traditional image-processing method. A second group received a slightly slower scan, also using the traditional method. The third group received the ultra-slow scan, but this group was split in two: half were processed with the traditional method, and the other half were processed with the new deep-learning computer program. The researchers then measured everything from the amount of radiation each patient received to the clarity of the final images. They looked at how much noise was in the pictures, how well the images showed the contrast between different tissues, and how often the distracting streaks appeared. Two experienced radiologists, who did not know which group each patient belonged to, also graded the images on a scale of one to five based on how easy they were to read and how clear the details were.

The results showed that the combination of the ultra-slow scan and the deep-learning program was the clear winner. This specific group received the lowest amount of radiation of all, with an average dose of 9.88 units, compared to higher doses in the other groups. More importantly, the images they produced were significantly clearer. The deep-learning program successfully removed the grainy noise that usually plagues low-dose scans, making the images much sharper. The researchers found that the noise in these images was reduced by more than half compared to the standard scans. The streaks and distortions that typically obscure the bowel walls were also dramatically suppressed. In the groups using the standard or slightly slower speeds, these artifacts appeared in the majority of scans, often making parts of the image difficult to interpret. In contrast, the group with the ultra-slow scan and deep-learning program showed these artifacts in only a small fraction of cases. The radiologists rated these images as excellent, noting that the bowel walls and surrounding structures were visible with exceptional clarity.

The study suggests that this dual approach works by attacking the problem from two sides. First, the slow scan speed prevents the artifacts from forming in the first place by ensuring the machine captures enough data. Second, the deep-learning program cleans up any remaining imperfections, distinguishing between true body structures and digital noise. This allowed the team to lower the radiation dose significantly without sacrificing image quality. The researchers noted that this method is particularly useful for patients who need repeated scans over time, such as those monitoring chronic bowel diseases, because it minimizes the total radiation they receive over their lifetime. However, they also pointed out a practical limitation: because the scanner moves so slowly, the scan takes longer to complete. The average time for this ultra-slow scan was nearly nine seconds, which is more than double the time of a standard scan. This requires patients to hold their breath for a longer period, which might be difficult for those who are very ill, in pain, or unable to follow breathing instructions.

Despite the longer scan time, the findings offer a promising path forward for abdominal imaging. The researchers concluded that this strategy effectively suppresses the streaks that usually ruin low-dose images while keeping radiation exposure low. It provides a way to see the intricate details of the digestive tract without the usual compromise between safety and clarity. While the study was conducted at a single hospital and focused specifically on the hollow organs of the abdomen, the results indicate that this method could become a standard way to perform these scans. By combining a simple adjustment to the scanner's speed with advanced computer processing, doctors may soon be able to provide safer, clearer diagnoses for patients who need them most.