In comparison to clinical-radiologic assessment, a deep learning CT radiomics nomogram had a 10 percent higher AUC and a 27 percent higher specificity for predicting the macrotrabecular massive subtype of hepatocellular carcinoma in external data testing.
An aggressive phenotype, the macrotrabecular-massive (MTM) subtype of hepatocellular carcinoma (HCC) carries high metastatic risk. While current imaging modalities may reveal independent predictive factors for MTM such as substantial necrosis and severe ischemia, these markers are reportedly plagued by low sensitivity and specificity for MTM.
However, emerging research suggests the use of a dual-energy computed tomography (DECT) radiomics nomogram may significantly improve risk assessment for MTM.
For the retrospective multicenter study, recently published in Radiology, researchers reviewed data from 262 patients with HCC and compared the DECT radiomics model with a clinical-radiologic model, which included factors such as an ⍺-fetoprotein level greater than 100 ng/mL, a platelet count greater than 100 x 109/L, and substantial necrosis.
The researchers found that the DECT radiomics nomogram in comparison to the clinical-radiologic model, had a 14 percent higher area under the receiver operating characteristic curve (AUC) for predicting MTM in the training set of 146 patients (91 percent vs. 77 percent) and a 10 percent higher AUC in the external test set of 81 patients (89 percent vs. 79 percent). While the DECT radiomics nomogram had lower sensitivity than the clinical-radiologic model (80 percent vs. 87 percent in the external data set, it had a 27 percent higher specificity rate (90 percent vs. 63 percent), according to the study authors.
“In this multicenter study, we found that an ⍺-fetoprotein level of more than 100 ng/mL and deep learning (DL) radiomics models of arterial phase (AP) virtual monoenergetic images, AP electron density, and portal venous phase iodine density were independent predictors for identifying MTM. These four predictors were included in a DL radiomics nomogram model, which showed better performance for predicting MTM HCC in our training, internal test, and external test data sets compared with a clinical-radiologic model … ,” noted Jin Wang, M.D., who is affiliated with the Department of Radiology at the Third Affiliated Hospital of Sun Yat-Sen University in Guangdong, China, and colleagues.
The study authors also pointed out that neither tumor size nor the presence of substantial necrosis affected the performance of the DECT radiomics nomogram.
Follow-up data for patients in the external data set revealed that patients who were not predicted to develop MTM, based on the DECT radiomics model, had a mean 36-month recurrence-free survival in comparison to a mean of 25 months for those predicted to be positive for MTM.
“The predicted positive MTM groups based on the DL radiomics nomogram had significantly shorter recurrence-free survival than predicted negative MTM groups in our three data sets,” emphasized Wang and colleagues.
(Editor’s note: For related content, see “Microvascular Invasion in Hepatocellular Carcinoma: Can a CT Radiomics Model Have an Impact?,” “Current Perspectives on the BCLC Staging System for Hepatocellular Carcinoma” and “Is Follow-Up Pelvic CT Coverage Necessary for Patients Treated for Hepatocellular Carcinoma?”)
In regard to study limitations, the researchers acknowledged that radiomics features may have been affected by differences in scanning protocols at different centers. They conceded the possibility of statistical bias in a post-hoc subgroup analysis due to unbalanced sample sizes of subgroups. The study authors also emphasized that large multicenter studies are necessary to assess the peritumoral deep learning radiomics model for detecting MTM.
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.