Modeling of pre-transplantation liver viability with spatial-temporal smooth variable selection.

Area of Science:

• Medical imaging
• Surgical innovation
• Biomedical engineering

Background:

• Accurate liver viability assessment is crucial for successful liver transplantation.
• Current methods are subjective, invasive, or yield inconsistent results.
• Donor liver utilization is limited by imprecise viability assessment.

Purpose of the Study:

• To develop a non-invasive statistical model for predicting liver viability.
• To utilize spatial-temporal infrared (IR) imaging data for this prediction.
• To enable binary prediction (acceptable/unacceptable) during organ preservation.

Main Methods:

• Monitoring liver surface temperature using IR thermography.
• Applying a spatial-temporal smooth variable selection (STSVS) method.
• Defining parameter smoothness across liver regions and time.

Main Results:

• The STSVS method demonstrated superior prediction performance over existing models (GLM, SVM, LASSO, Fused LASSO).
• Case studies using porcine livers validated the STSVS method's efficacy.
• Identified key predictors, highlighting the importance of lobe edges in viability prediction.

Conclusions:

• The proposed STSVS method offers the best performance for liver viability prediction.
• This real-time, non-damaging approach can increase donor liver utilization.
• It overcomes limitations of time-consuming and imprecise traditional assessments.