An MRI-Based Deep Transfer Learning Radiomics Nomogram to Predict Ki-67 Proliferation Index of Meningioma

Area of Science:

• Neuro-oncology
• Medical imaging analysis
• Computational pathology

Background:

• Meningiomas are the most common primary brain tumors.
• Accurate prediction of Ki-67 proliferation index is crucial for grading and treatment planning.
• Current methods for Ki-67 assessment can be subjective and time-consuming.

Purpose of the Study:

• To develop and validate a nomogram for predicting the Ki-67 proliferation index in meningioma.
• To integrate clinical, radiomics, and deep transfer learning (DTL) features for improved prediction accuracy.
• To evaluate the performance of the developed nomogram against traditional models.

Main Methods:

• Retrospective analysis of 318 meningioma cases.
• Extraction and selection of clinical, radiomics, and DTL features.
• Construction of individual models (clinical, radiomics, DTL) and a combined Deep Transfer Learning Radiomics (DTLR) nomogram.
• Performance evaluation using Area Under the Receiver Operator Characteristic Curve (AUC), accuracy, sensitivity, and specificity.
• Comparison of models using Delong test and Decision Curve Analysis (DCA).

Main Results:

• The DTLR nomogram achieved the highest AUC of 0.779, outperforming individual clinical (0.746), radiomics (0.75), and DTL (0.717) models.
• The DTLR nomogram demonstrated good accuracy (0.734), sensitivity (0.719), and specificity (0.75) in the test set.
• Decision Curve Analysis indicated that the DTLR nomogram provided greater net benefit across various threshold probabilities.

Conclusions:

• The DTLR nomogram is a robust and accurate tool for predicting the Ki-67 proliferation index in meningioma.
• This integrated approach offers a valuable, objective method for meningioma evaluation.
• The DTLR nomogram has the potential to aid in clinical decision-making and improve patient management.