Radiomic Analysis for Ki-67 Classification in Small Bowel Neuroendocrine Tumors.

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Small bowel neuroendocrine tumors (NETs) are a heterogeneous group of neoplasms.
• Assessing the proliferative rate, often indicated by Ki-67 expression, is crucial for prognosis and treatment planning.
• Current methods for determining Ki-67 expression are invasive and may not represent the entire tumor heterogeneity.

Purpose of the Study:

• To analyze radiomic features from CT images of small bowel NETs.
• To evaluate the association between these radiomic features and Ki-67 expression.
• To explore the potential of radiomics for non-invasively predicting tumor proliferative rate.

Main Methods:

• 128 small bowel NET lesions from 34 patients underwent manual segmentation on CT images.
• 107 radiomic features were extracted using PyRadiomics, with correlation filtering reducing the set to 41.
• Statistical analyses identified 19 significant texture-related features, and machine learning models (including Random Forest) were trained and validated.

Main Results:

• Random Forest model achieved the highest performance with an AUC of 0.80 and F1 score of 0.813.
• The model demonstrated a low false negative rate (15.3%), indicating potential for accurate identification of aggressive lesions.
• Texture-based radiomic features were predominantly associated with Ki-67 expression.

Conclusions:

• Radiomics shows promise for non-invasively assessing the proliferative rate of small bowel NETs.
• This approach can identify patterns linked to histopathological markers like Ki-67.
• Further research is warranted to explore the clinical utility of radiomics in NET management.