SIR 2024
Interventional Oncology
Mohammad Mirza-Aghazadeh-Attari, MD, MPH
Postdoctoral Research Fellow
Johns Hopkins University
Financial relationships: Full list of relationships is listed on the CME information page.
Tara Srinivas, MS
Medical Student
Johns Hopkins School of Medicine
Disclosure information not submitted.
Alan J. Kim, BS (he/him/his)
Medical Student
Johns Hopkins University School of Medicine
Financial relationships: Full list of relationships is listed on the CME information page.
Arun Kamireddy, MD, MBBS
Research fellow
Johns Hopkins University School of Medicine
Financial relationships: Full list of relationships is listed on the CME information page.
Clifford R. Weiss, MD, FSIR
Professor of Radiology and Biomedical Engineering
The Johns Hopkins Hospital
Financial relationships: Full list of relationships is listed on the CME information page.
To assess the prognostic value of radiomics features extracted from preprocedural imaging in predicting early recurrence and tumoral progression (defined as recurrence within 2 years) after ablative therapy.
Materials and methods:
A systematic search was conducted in Scopus, Web of Science, PubMed, and Embase to retrieve relevant studies. Inclusion criteria required studies to report the diagnostic accuracy of radiomics features for predicting the recurrence of very early and early stage HCC within two years following ablative procedures, including radiofrequency ablation, microwave ablation, and cryoablation. The pooled diagnostic accuracy of the radiomics models was determined using a random-effects model. Heterogeneity was assessed using the Q statistic and I2 index. Publication bias was evaluated using Deek’s Funnel plot. All statistical analyses were performed using Stata (Version 18) and the open-source R statistical software. The Radiomics Quality Score Version 2 was employed to assess methodological quality of the studies.
Results:
A total of 18 databases from 11 studies (6 MRI, 3 Ultrasound, and 2 CT) encompassing 1786 patients with HCC were included. Deek’s plot indicated no significant publication bias (P=0.4). Sixteen databases from 10 studies reported results for recurrence in a 2-year follow-up period. The pre-test probability of recurrence within 2 years was 49%. The pooled sensitivity, specificity, and area under the ROC curve were 0.738 (95% CI: 0.66-0.79), 0.765 (95% CI: 0.66-0.83), and 0.81 (95% CI: 0.77-0.84), respectively. Four studies, encompassing 8 datasets, reported results for a 1-year follow-up, with cumulative sensitivity and specificity of 0.9 (95% CI: 0.78-0.95) and 0.68 (95% CI: 0.62-0.74), respectively. The mean radiomics quality score of the studies was 23 ± 6 (34 ± 9%), indicating an overall low quality of the studies. None of the studies adhered to specific principles, such as cost-effectiveness evaluation, validation of radiomics pipelines using phantoms, or integration into clinical practice.
Conclusion:
Automated models based on radiomics features show good sensitivity and specificity in predicting early recurrence in early-stage HCC patients undergoing curative ablative therapy. The principal factor limiting the generalizability of the results is the significant heterogeneity observed in the methodologies and the low quality of the methodologies used. Future studies should place special focus on multi-center validation and design their methodologies in complete adherence to quality measure checklists.