LINNDA: Lymphoma identification through neural network detection aid.

Area of Science:

• Neuro-oncology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Accurate preoperative differentiation between primary central nervous system lymphoma (PCNSL) and glioblastoma (GBM) is critical for effective patient management and surgical strategy.
• Distinguishing between PCNSL and GBM preoperatively presents a significant diagnostic challenge in neuro-oncology.

Purpose of the Study:

• To assess the diagnostic performance of an AI-assisted workflow, LINNDA (lymphoma identification through neural network detection aid), in differentiating PCNSL from GBM.
• To compare the diagnostic accuracy of the AI-assisted workflow against human expert consensus and individual raters.

Main Methods:

• Ten clinicians independently reviewed 46 cases of GBM and PCNSL.
• An AI workflow (LINNDA) integrated a DenseNet169 neural network as a "tie-breaker" when clinicians disagreed.
• The AI's diagnostic performance was evaluated against human expert consensus and in comparison to a third human rater across numerous scenarios.

Main Results:

• Integrating AI predictions into the diagnostic process improved overall accuracy to 89.9%, surpassing expert consensus.
• The AI-assisted approach demonstrated superiority over a third human rater in 5,108 evaluated scenarios.
• LINNDA achieved a negative predictive value of 97% for ruling out PCNSL, supporting its clinical utility.

Conclusions:

• The AI-assisted LINNDA workflow significantly enhances the accuracy of preoperative differentiation between PCNSL and GBM.
• LINNDA provides a reliable tool for clinical decision-making, particularly in cases with diagnostic uncertainty.
• This AI approach offers a valuable adjunct to human expertise in neuro-oncology diagnostics.