Clinical Scorecard: Utilizing Deep Learning to Evaluate Survival Outcomes in Hepatocellular Carcinoma Patients Receiving Intra-Arterial Therapies Based on Proliferative Subtype

At a Glance

Key Highlights

• Proliferative HCC subtype is linked to aggressive disease and poor prognosis but is challenging to identify non-invasively.
• A novel deep learning model (Prototype Mamba Net) using CT imaging achieved high accuracy (AUC ~0.79–0.83) in detecting proliferative HCC.
• Prognostic nomograms combining radiomic and clinical data outperformed traditional staging systems in survival prediction and informed personalized intra-arterial therapy choices.

Guideline-Based Recommendations

Diagnosis

• Utilize contrast-enhanced CT imaging features and deep learning models to non-invasively identify proliferative HCC subtype.
• Recognize imaging predictors such as lobulated tumor margins, satellite nodules, mosaic architecture, and rim-type arterial phase hyperenhancement.

Management

• Consider hepatic arterial infusion chemotherapy (HAIC) over transarterial chemoembolization (TACE) for high-risk proliferative HCC patients to improve survival.
• Integrate deep learning-based phenotyping to guide personalized intra-arterial therapy selection in unresectable HCC.

Monitoring & Follow-up

• Employ prognostic nomograms combining radiomic and clinical variables for ongoing survival risk assessment post-therapy.

Risks

• Recognize that proliferative HCC is associated with poorer outcomes after TACE or surgical resection compared to non-proliferative subtypes.
• Be aware of heterogeneity in tumor biology affecting treatment response and survival.

Patient & Prescribing Data

Treatment-naïve patients with unresectable hepatocellular carcinoma undergoing intra-arterial therapies

Among high-risk proliferative HCC patients, HAIC demonstrated a significant survival benefit compared to TACE; no significant survival difference was observed in low-risk patients.

Clinical Best Practices

• Incorporate advanced deep learning models analyzing CT imaging to non-invasively subtype HCC before treatment initiation.
• Use combined radiomic and clinical data to improve prognostic accuracy beyond traditional staging systems.
• Tailor intra-arterial therapy choice (HAIC vs. TACE) based on proliferative subtype risk stratification to optimize patient outcomes.

References

• Bao et al. CT-based predictors of proliferative HCC
• Kang et al. MRI features linked to proliferative HCC
• Phase III studies comparing HAIC and TACE
• Vaswani et al. Transformer-based deep learning models