decision-tree

decision-tree implements a decision-tree methodology that classifies WHO categories of B-cell chronic lymphoproliferative disorders using multiparametric flow cytometry immunophenotypic data.

Key Features:

• Decision-Tree Structure: Employs a decision tree composed of logistic function nodes that branch into sets representing potential distinct diagnoses.
• Input Data: Processes multiparametric flow cytometry data from peripheral blood and bone marrow for immunophenotypic characterization.
• Regularization (Lasso): Applies Lasso regularization to logistic nodes to mitigate overfitting and enhance model generalizability.
• Transparent Binary Nodes: Uses up to seven binary transparent decision nodes to provide interpretable diagnostic pathways.
• Validation and Performance: Validated on diagnostic samples from 283 patients with mature lymphoid leukemias/lymphomas, achieving 95% correctness in cross-validation and 100% in the training sample, with single diagnoses for 61% and multiple possible diagnoses for 34% of cases.

Scientific Applications:

• Differential Diagnosis: Distinguishes between WHO categories of B-cell chronic lymphoproliferative disorders based on flow cytometry immunophenotypes.
• Research and Development: Provides a validated computational framework to support research into leukemia and lymphoma diagnostic criteria and therapeutic strategies.

Methodology:

Constructs a decision tree of logistic regression nodes with Lasso regularization and up to seven binary decision nodes.