Artificial intelligence in child abuse imaging.

There have been rapid advances in artificial intelligence (AI) technology in recent years, and the field of diagnostic imaging is no exception. Just as digital technology revolutionized how radiology is practiced, so these new technologies also appear poised to bring sweeping change. As AI tools make the transition from the theoretical to the everyday, important decisions need to be made about how they will be applied and what their role will be in the practice of radiology. Pediatric radiology presents distinct challenges and opportunities for the application of these tools, and in this article we discuss some of these, specifically as they relate to the prediction, identification and investigation of child abuse.

Comparison of Standardized Uptake Values in Normal Structures Between PET/CT and PET/MRI in a Tertiary Pediatric Hospital: A Prospective Study.

OBJECTIVE: The purpose of this study was to compare standardized uptake values (SUVs) of normal tissues using MR attenuation-corrected versus CT attenuation-corrected (18)F-FDG PET in a pediatric population. SUBJECTS AND METHODS: Thirty-five patients (21 boys; mean age, 13.3 years) referred for 47 PET/CT scans were recruited to undergo PET/MRI. MR attenuation correction was performed using an automated three-segment model. ROIs were drawn over nine normal structures to estimate SUV(min), SUV(mean), and SUV(max). Pearson rank correlation coefficients were calculated to compare SUVs obtained from MR and CT attenuation correction. In nine patients who underwent multiple PET/MRI studies, coefficients of variance and intraclass correlation coefficients were calculated to evaluate intrapatient SUV(max) variation. RESULTS: Mean (± SD) time to imaging after FDG injection was 108 ± 17 minutes for PET/CT and 61 ± 6 minutes for PET/MRI. PET/MRI SUVs in all tissues were lower than those for PET/CT (mean difference, -28.9% ± 31.1%; p < 0.05). Very high or high correlation between PET/MRI and PET/CT SUV(max) was found in brain (r = 0.72), myocardium (r = 0.95), and bone marrow (r = 0.85) (p < 0.001). Moderate correlation was found in liver (r = 0.54), fat (r = 0.41), mean blood pool (r = 0.40), and psoas muscle (r = 0.38) (p < 0.01). Weak correlation was found in lung (r = 0.12) and iliacus muscle (r = 0.12). Compared with PET/CT, PET/MRI systematically undermeasured SUV. In nine patients who underwent multiple PET/MRI examinations, moderate or strong agreement was found in the SUV(max) of six of nine tissues, similar to the corresponding PET/CT examinations. CONCLUSION: Our study showed overall high correlation for SUV measurements obtained from MR attenuation correction compared with CT attenuation correction, although PET/MRI underestimated SUV compared with PET/CT. SUVs measured from PET/MRI indicated good intrapatient reliability.

Qualitative FDG PET Image Assessment Using Automated Three-Segment MR Attenuation Correction Versus CT Attenuation Correction in a Tertiary Pediatric Hospital: A Prospective Study.

OBJECTIVE: The purpose of this study was to systematically evaluate the diagnostic quality of (18)F-FDG PET images generated using MR attenuation correction (MRAC) compared with those images generated using CT attenuation correction (CTAC) in a pediatric population. SUBJECTS AND METHODS: Forty-two patients (mean age, 12.8 years; percentage who were male, 57%) who were referred for 62 indicated whole-body PET/CT studies were prospectively recruited to undergo PET/MRI examinations during the same clinic visit in which PET/CT was performed. MRAC was performed using an automatic three-segment model. Three nuclear radiologists scored the diagnostic quality of the PET images generated by MRAC and CTAC using a Likert scale (range of scores, 1-5). Images graded with a score of 1-3 were considered clinically unacceptable, whereas images with a score of 4-5 were considered clinically acceptable. A Wilcoxon signed-rank test was used to compare differences in the grading of PET/MRI and PET/CT images. The Fisher exact test was used to evaluate potential differences in clinically acceptable image quality and the presence of artifact. Fleiss kappa statistics were used to examine interobserver agreement. RESULTS: There was no statistically significant difference in the proportion of PET images generated with MRAC and CTAC for which image quality was considered clinically acceptable. A total of 3.9% of PET assessments generated with MRAC were of unacceptable image quality, compared with 2.2% of PET images generated with CTAC. Two of the three radiologists who reviewed the PET images reported the presence of artifacts more often on MRAC-derived images, and they graded the mean quality of these images 0.48 and 0.29 points lower on the 5-point Likert scale than they graded the mean quality of CTAC-derived images (p < 0.0001). Interobserver agreement was fair (κ = 0.39). CONCLUSION: The diagnostic quality of PET images obtained from a pediatric population with the use of an automatic three-segmentation MRAC method was comparable to that of PET images obtained with the use of CTAC.

Lung cancer imaging.

Lung cancer remains the leading cause of cancer-related deaths in the US. Imaging plays an important role in the diagnosis, staging, and follow-up evaluation of patients with lung cancer. With recent advances in technology, it is important to update and standardize the radiological practices in lung cancer evaluation. In this article, the authors review the main clinical applications of different imaging modalities and the most common radiological presentations of lung cancer.