Radiomics-driven spectral profiling of six kidney stone types with monoenergetic CT reconstructions in...

Area of Science:

• Urology
• Medical Imaging
• Radiology

Background:

• Urolithiasis (kidney stones) diagnosis and management are critical for patient care.
• Accurate differentiation of kidney stone subtypes is essential for effective, personalized treatment strategies.
• Conventional imaging techniques like CT and DECT have limitations in detailed stone characterization.

Purpose of the Study:

• To investigate the efficacy of photon-counting computed tomography (PCCT) combined with radiomics for automated kidney stone characterization.
• To develop and evaluate a novel method for detailed kidney stone classification beyond current imaging capabilities.
• To enhance diagnostic accuracy and inform personalized therapeutic approaches for urolithiasis.

Main Methods:

• An ex vivo study involving 135 kidney stones classified by infrared spectroscopy.
• Kidney stones were scanned using PCCT, generating various monoenergetic reconstructions.
• Radiomics features were extracted and analyzed using Random Forest classifiers, including a combined spectral model.

Main Results:

• The combined radiomics model achieved an AUC of 0.95 and a test accuracy of 0.81 in differentiating six kidney stone types.
• This spectral-driven radiomics approach significantly outperformed individual monoenergetic reconstructions and SPP parameters.
• Key radiomics features identified included NGTDM_Strength and wavelet-LLH_firstorder_Variance.

Conclusions:

• Radiomics analysis of PCCT data significantly improves the differentiation of kidney stone subtypes.
• The combined spectral radiomics model demonstrates high accuracy for non-invasive kidney stone characterization.
• This PCCT-based imaging strategy offers potential for optimizing personalized treatment and reducing stone recurrence.