Using AI to Boost Chest Disease Diagnosis with Thick-Slice CT

CT scans are crucial for spotting chest diseases. The quality of these images largely depends on their spatial resolution. Thick-slice CT, which is still commonly used due to cost concerns, has a lower spatial resolution that can make diagnoses tricky. Researchers recently developed a deep learning model that can turn thick-slice CT scans into thin-slice scans. This model was tested on a large group of participants and showed promising results. The quality of the synthetic thin-slice CT images was comparable to real thin-slice CT images. Four radiologists found that diagnosing diseases like community-acquired pneumonia was more accurate with synthetic thin-slice CT than with thick-slice CT, and the results were just as good as with real thin-slice CT. For detecting lung nodules, synthetic thin-slice CT did even better than thick-slice CT and matched the performance of real thin-slice CT. This suggests that this AI model could be a practical alternative when high-quality thin-slice CT scans are needed but not available.