A 49-study meta-analysis revealed that MRI-based and CT-based radiomics models outperformed clinical models for predicting recurrence of hepatocellular carcinoma.
Radiomics models based on magnetic resonance imaging (MRI) or computed tomography offer a superior concordance index (C-index), sensitivity and specificity than clinical models for predicting hepatocellular carcinoma (HCC), according to a new meta-analysis.
For the meta-analysis, recently published in Academic Radiology, researchers reviewed data from 49 studies that involved the use of radiomics-based machine learning models for predicting recurrence in patients who had been pathologically or clinically diagnosed with HCC. The reviewed studies included 26 CT-based radiomics models, 19 MRI-based radiomics models and 4 radiomics models based on ultrasound or CEUS, according to the study. The study authors compared the C-index, sensitivity rates and specificity rates for the radiomics models versus clinical models, which were included in 33 of the reviewed studies.
The researchers found that the CT-based radiomics model had a 74.7 percent C-index, 72 percent sensitivity and 77 percent specificity in the validation cohort. In validation testing, the MRI-based radiomics model offered a 78.8 percent C-index, 74 percent sensitivity and 78 percent specificity. In contrast, the study authors noted that validation testing with clinical models revealed a 67.1 percent C-index, 66 percent sensitivity and 71 percent specificity for predicting HCC recurrence.
In comparison to the clinical model alone, the study authors also noted that combining the MRI-based radiomics model with the clinical model in a validation cohort resulted in a greater than 15 percent improvement in the C-index (82.6 percent), a 16 percent improvement in sensitivity (82 percent) and an 11 percent improvement in specificity (82 percent) for predicting HCC recurrence.
The researchers also noted differences between the clinical model alone versus the combination of the CT-based radiomics model and clinical model in validation testing, citing a nearly 12 percent C-index improvement (79 percent), a 5 percent increase in sensitivity (71 percent) and a 7 percent increase in specificity (78 percent).
“The analysis results suggest that the radiomics signature has the potential to be a useful tool for predicting HCC recurrence. Furthermore, it is found that combined models integrating clinical factors and radiomics signatures (regardless of the modality used) perform better compared to those using clinical variables alone,” wrote study co-author Pin-Tong Huang, M.D., a professor, chief physician and chair of the Ultrasound Department at the Second Affiliated Hospital of Zhejiang University School of Medicine in Hangzhou, China, and colleagues.
The results were more mixed with ultrasound-based radiomics models, according to the researchers. In the validation cohort, the CEUS-based radiomics model had a 76.3 percent C-index, sensitivity ranging between 58 and 90 percent, and specificity ranging between 58 and 96 percent. In the validation cohort of one study, an ultrasound-based radiomics model offered a 56 percent C-index, 35 percent sensitivity and 96 percent specificity.
While noting the limited number of ultrasound and CEUS studies in the meta-analysis and conceding that ultrasound-based radiomics models “exhibit certain limitations,” Huang and colleagues noted improved HCC recurrence prediction with combined ultrasound-based radiomics and clinical models. In a validation cohort, the combined ultrasound model exhibited a 76 percent C-index, 69 percent sensitivity and 93 percent specificity.
“Our meta-analysis indicates that the radiomics model based solely on ultrasound images is less adept at predicting HCC recurrence when compared to the model based on clinical characteristics, whereas the combined model incorporating (ultrasound)-based radiomics and clinical features outperforms the model grounded solely in clinical features,” added Huang and colleagues.
(Editor’s note: For related content, see ”Can a CT Radiomics Model Improve Risk Assessment for Macrotrabecular-Massive Hepatocellular Carcinoma?,” “Microvascular Invasion in Hepatocellular Carcinoma: Can a CT Radiomics Model Have an Impact?” and “Current Perspectives on the BCLC Staging System for Hepatocellular Carcinoma.”)
In regard to study limitations, the study authors noted the majority of the reviewed studies involved single-center research conducted in China, which may limit extrapolation of meta-analysis findings to a broader population. The researchers also conceded a degree of heterogeneity with respect to the modeling methods, feature extract and imaging modality utilized in the reviewed studies.
Leading Breast Radiologists Discuss the USPSTF Breast Cancer Screening Recommendations
May 17th 2024In recognition of National Women’s Health Week, Dana Bonaminio, MD, Amy Patel, MD, and Stacy Smith-Foley, MD, shared their thoughts and perspectives on the recently updated breast cancer screening recommendations from the United States Preventive Services Task Force (USPSTF).
Multicenter CT Study Shows Benefits of Emerging Diagnostic Model for Clear Cell Renal Cell Carcinoma
May 15th 2024Combining clinical and CT features, adjunctive use of a classification and regression tree (CART) diagnostic model demonstrated AUCs for detecting clear cell renal cell carcinoma (ccRCC) that were 15 to 22 percent higher than unassisted radiologist assessments.
CT Study: AI Algorithm Comparable to Radiologists in Differentiating Small Renal Masses
May 14th 2024An emerging deep learning algorithm had a lower AUC and sensitivity than urological radiologists for differentiating between small renal masses on computed tomography (CT) scans but had a 21 percent higher sensitivity rate than non-urological radiologists, according to new research.