Overall survival and toxicity of hepatocellular carcinoma Barcelona Clinic Liver Cancer B patients receiving Y90 radioembolization: analysis of the radiation-emitting SIR-spheres in non-resectable liver tumor (RESiN) registry.

Background: To evaluate overall survival (OS), progression-free survival (PFS) and toxicity after resin Yttrium-90 (Y-90) radioembolization in Barcelona Clinic Liver Cancer B (BCLC B) hepatocellular carcinoma (HCC) patients using the Bolondi subgroup classification. Methods: A total of 144 BCLC B patients were treated between 2015-2020. Patients were broken into 4 subgroups by tumor burden/liver function tests with 54, 59, 8 and 23 in subgroups 1, 2, 3 and 4. OS and PFS were calculated with Kaplan-Meier analysis with 95% confidence intervals. Toxicities were assessed using Common Terminology Criteria for Adverse Events (CTCAE) v5. Results: Prior resection and chemoembolization were performed in 19 (13%) and 34 (24%) of patients. There were no deaths within 30 days. Median OS and PFS for the cohort were 21.5 and 12.4 months. Median OS was not reached for subgroup 1 at a mean 28.8 months, and was 24.9, 11.0 and 14.6 months for subgroups 2-4 (χ2=19.8, P=0.0002). PFS by BCLC B subgroup was 13.8, 12.4, 4.5, and 6.6 months (χ2=16.8, P=0.0008). The most common Grade 3 or 4 toxicities were elevated bilirubin (n=16, 13.3%) and decreased albumin (n=15, 12.5%). Grade 3 or greater bilirubin (32% vs. 10%, P=0.03) and albumin (26% vs. 10%, P=0.03) toxicity were more common in the subgroup 4 patients. Conclusions: The Bolondi subgroup classification stratifies OS, PFS and development of toxicity in patients treated with resin Y-90 microspheres. OS in subgroup 1 approaches 2.5 years and Grade 3 or greater hepatic toxicity profile in subgroups 1-3 is low.

Six-Year Evaluation of Same-Day Discharge following Conventional Transarterial Chemoembolization of Hepatocellular Carcinoma.

PURPOSE: To evaluate whether same-day discharge increased the incidence of 30-day readmission (30dR) after conventional transarterial chemoembolization (TACE) for hepatocellular carcinoma (HCC) at a single institution. MATERIALS AND METHODS: In this retrospective study, 253 patients with HCC underwent 521 transarterial chemoembolization procedures between 2013 and 2020. TACE was performed with 50-mg doxorubicin/10-mg mitomycin C/5-10-mL ethiodized oil/particles. Patients not requiring intravenous pain medications were discharged after a 3-hour observation, and 30dR was tracked. The primary objective was to determine the incidence of 30dR in same-day discharge patients versus patients admitted for observation using the chi-square test. Secondary objectives assessed factors associated with overnight admission and factors predictive of 30dR using generalized estimated equation calculations and logistic regression. RESULTS: In the cohort, 24 readmissions occurred within 30 days (4.6%). Same-day discharge was completed after 331 TACE procedures with sixteen 30dRs (4.8%). Patients admitted overnight were readmitted 8 times after 190 TACE procedures (4.2%). This difference was not statistically significant (P = .4). Factors predicting overnight admission included female sex (58/190 [30.5%] vs 58/331 [17.5%], P < .001) and tumor size of ≥3.8 cm (104/190 [55%] vs 85/190 [45%]). Factors predicting 30dR included female sex (10/116 [8.6%] vs 14/405 [0.2%]) and younger age (median [interquartile range], 63 years [55-65 years] vs 65 years [59-71 years]). At regression, factors predictive of 30dR were Child-Pugh Class B/C (odds ratio [OR], 2.1; P = .04) and female sex (OR, 2.9; P = .004). CONCLUSIONS: Same-day discharge after conventional TACE is a safe and effective strategy with 30dR rate of <5%, similar to overnight observation.

Effectiveness of Deep Learning Algorithms to Determine Laterality in Radiographs.

Develop a highly accurate deep learning model to reliably classify radiographs by laterality. Digital Imaging and Communications in Medicine (DICOM) data for nine body parts was extracted retrospectively. Laterality was determined directly if encoded properly or inferred using other elements. Curation confirmed categorization and identified inaccurate labels due to human error. Augmentation enriched training data to semi-equilibrate classes. Classification and object detection models were developed on a dedicated workstation and tested on novel images. Receiver operating characteristic (ROC) curves, sensitivity, specificity, and accuracy were calculated. Study-level accuracy was determined and both were compared to human performance. An ensemble model was tested for the rigorous use-case of automatically classifying exams retrospectively. The final classification model identified novel images with an ROC area under the curve (AUC) of 0.999, improving on previous work and comparable to human performance. A similar ROC curve was observed for per-study analysis with AUC of 0.999. The object detection model classified images with accuracy of 99% or greater at both image and study level. Confidence scores allow adjustment of sensitivity and specificity as needed; the ensemble model designed for the highly specific use-case of automatically classifying exams was comparable and arguably better than human performance demonstrating 99% accuracy with 1% of exams unchanged and no incorrect classification. Deep learning models can classify radiographs by laterality with high accuracy and may be applied in a variety of settings that could improve patient safety and radiologist satisfaction. Rigorous use-cases requiring high specificity are achievable.