New prospective research demonstrated that the combination of ultra-low-dose (ULD) computed tomography (CT) and an artificial intelligence (AI)-based denoising method offered a greater than 95 percent sensitivity rate for differentiating negative and actionable Lung-RADS category findings with significantly reduced radiation dosing.
Could artificial intelligence (AI) allow a significant reduction in radiation dosing for computed tomography (CT) without sacrificing the detection of actionable Lung-RADS findings?
In a new prospective study, recently published in Academic Radiology, researchers examined the use of an emerging denoising technique in combination with ultra-low-dose (ULD) CT for lung cancer screening in 123 patients (mean age of 62.6). The AI-based denoising method utilized a fully convolutional network, trained with stacked convolutional denoising auto-encoders, which allowed feature extraction for perceptual loss computation, according to the study authors.
The researchers reported that the overall average effective radiation dose for low-dose (LD) CT was 1.01 + 0.06 mSv in comparison to 0.24 + 0.08 mSv for ULD CT, equating to a 76 percent reduction in the average radiation dose.
The study authors also found no significant differences between the sensitivity rates and accuracy rates for dULD (denoising method for ULD) and ULD in differentiating between negative and actionable Lung-RADS findings. In comparing dULD vs. LD, the researchers noted sensitivity rates ranging between 95.5 to 98.1 percent and accuracy rates ranging between 95 to 97.5 percent. In looking at ULD vs. LD, the study authors found sensitivity rates ranging from 94.6 to 97.2 percent and accuracy rates ranging between 94.6 to 97.2 percent.
“Although we found no statistically significant difference between Lung-RADS results of LD vs. ULD and LD vs. dULD, the dULD images are not as noisy, easier on the eyes and similar to the lD images radiologists are used to,” wrote Edith Michelle Marom, M.D., a professor at the University of Texas MD Anderson Cancer Center. "It was also possible to answer the clinical questions and to identify ancillary findings which are of medical importance.”
(Editor’s note: For related content, see “Deep Learning Model May Predict Lung Cancer Risk from a Single CT Scan,” “Nine Takeaways from Recent Meta-Analysis on Lung Cancer Screening with Low-Dose CT” and “Seven Takeaways from Lung-RADS 2022 for CT Lung Cancer Screening.”)
For coronary artery calcifications (CAC), the researchers noted that dULD had sensitivity rates ranging between 93.9 to 97.6 percent and an accuracy rate of 91.7 percent. The study authors also said the use of dULD was beneficial for diagnosing subsolid nodules.
“The importance in correctly identifying part solid nodules lies in the fact that they are more likely to be malignant. The solid component of a sub-solid nodule correlates with cancer invasion for small adenocarcinomas,” pointed out Marom and colleagues.
In addition to a low number of patients in the study cohort, the authors acknowledged the low sensitivity of dULD imaging for emphysema (ranging from 56.3 to 56.8 percent) was a key limitation of the study.
MRI-Based Deep Learning Algorithm Shows Comparable Detection of csPCa to Radiologists
May 8th 2024In a study involving over 1,000 visible prostate lesions on biparametric MRI, a deep learning algorithm detected 96 percent of clinically significant prostate cancer (csPCa) in comparison to a 98 percent detection rate for an expert genitourinary radiologist.
Study Finds High Concordance Between AI and Radiologists for Cervical Spine Fractures on CT
May 6th 2024Researchers found a 98.3 percent concordance between attending radiology reports and AI assessments for possible cervical spine fractures on CT, according to new research presented at the 2024 ARRS Annual Meeting.
Can a CT-Based Radiomics Model Bolster Detection of Malignant Thyroid Nodules?
May 3rd 2024A computed tomography (CT)-based radiomics model that includes 28 radiomic features showed significantly higher accuracy, sensitivity, and specificity than conventional CT in differentiating benign and malignant thyroid nodules, according to newly published research.
New Literature Review Finds ChatGPT Effective in Radiology in 84 Percent of Studies
April 29th 2024While noting a variety of pitfalls with the chatbot ranging from hallucinations to improper citations, the review authors found the use of ChatGPT in radiology demonstrated “high performance” in 37 out of 44 studies.