Computational sonomiography system for the rectus femoris muscle in DICOM images for morphological and textural quantification with spatial calibration

Abstract

Introduction

Morphological assessment of the rectus femoris muscle using B-mode ultrasound allows for the estimation of parameters related to muscle architecture, functional capacity, and changes associated with patient rehabilitation. However, manual measurement of anatomical cross-sectional area (ACSA), muscle thickness (MT), penneation angle (PA), and fascicle length (FL) can be time-consuming, operator-dependent, and poorly reproducible, especially when images are stored in DICOM format and do not always include complete spatial calibration metadata.

Objective

To develop a computational sonography technology to quantify morphological parameters and texture descriptors in longitudinal and cross-sectional DICOM images of the rectus femoris muscle, to support functional muscle assessment.

Methodology

A pipeline was designed in MATLAB/Python to interactively select longitudinal and transverse DICOM images from a directory, read DICOM metadata, choose frames in multi-frame studies, and calibrate the scale using the DICOM PixelSpacing metadata or a manual 1-cm reference visible in the B-mode ultrasound image. In the longitudinal plane, PA, MT, and FL were estimated using aponeurosis detection, local orientation analysis, and ray-casting; when the fascia did not fit within the field of view, FL was estimated using trigonometry as MT/sin (PA). In the transverse plane, ACSA was calculated from a user-defined or user-adjusted region of interest. Estimates of PCSA and FCSA were obtained from ACSA, PA, and FL. Additionally, texture features were extracted within the ROI, including first-order, GLCM, GLRLM, LBP, and Gabor filter bank. The system generates visual overlays and a single CSV file for data analysis.

Results

A reproducible workflow was established that integrates spatial calibration, anatomical measurement, functional estimation, and textural analysis into a single procedure. The design allows for the comparison of direct and trigonometric measurements of FL, documents the method used, visually verifies the aponeuroses and the transverse ROI, and consolidates ACSA, PCSA, FCSA, PA, MT, FL, and textures into an exportable database. As an operational result, the tool reduces scattered manual steps and facilitates quality review through overlays.

Conclusions

This proposal presents a sonomiography system designed for translational research in rehabilitation, with the potential to standardize the quantification of the rectus femoris muscle in B-mode ultrasound DICOM images, support studies on sarcopenia, functional recovery, and therapeutic follow-up, and generate morphological and textural variables useful for predictive models. The next phase will involve validation against expert measurements, intra- and inter-observer reliability analysis, and evaluation in clinical cohorts.