Recurrence-specific supervised graph clustering for subtyping Hodgkin Lymphoma radiomic phenotypes.

Area of Science:

• Oncology
• Medical Imaging
• Data Science

Background:

• Accurate prediction of therapy failure or recurrence in Hodgkin Lymphoma is crucial for personalized treatment strategies.
• Radiomics offers potential for quantitative prognostic factors from medical images, but faces limitations like high dimensionality.
• Existing radiomic studies require enhanced methods for comprehensive patient representation and robust risk stratification.

Purpose of the Study:

• To develop an advanced method for predicting Hodgkin Lymphoma recurrence risk at baseline.
• To overcome limitations of current radiomic approaches through an exhaustive patient representation.
• To stratify patients into distinct risk classes for tailored clinical management.

Main Methods:

• An exhaustive patient representation framework was employed.
• A recurrence-specific multi-view supervised clustering algorithm was developed.
• Patient-to-patient similarity graphs were estimated to learn recurrence probability.

Main Results:

• The proposed algorithm successfully estimated patient similarity and learned recurrence probability.
• Patients were stratified into two distinct risk classes based on predicted recurrence likelihood.
• Clinical variables were characterized for each identified risk group.

Conclusions:

• The developed multi-view supervised clustering algorithm shows promise for predicting Hodgkin Lymphoma recurrence.
• This approach can significantly impact clinical practice by identifying high-risk patients early.
• Further research can refine these methods for improved prognostic accuracy in lymphoma patients.