Applying Machine Learning for Medical Image Processing

2. Medical Imaging Modalities

Medical images are acquired through various physical principles and instrumentation:

X-ray: High contrast 2D projection images produced by differential absorption of x-ray photons passing through the body. Provides views of bones and calcifications. Low cost and wide availability have made it the most common modality. Based on differential absorption of ionizing radiation. Acquired via single exposure. Provides high contrast view of bony anatomy and pathology. Reader expertise needed for diagnosis (Khang & Rani et al., 2022).

CT: 3D volumetric data generated by tomography using multiple x-ray projections. Provides fine anatomic detail of both bone and soft tissue across the body. Captures multiple x-ray projections around patient using rotating gantry. Computational tomography reconstructs 3D volume. Provides detailed anatomies of both soft tissue and bone.

MRI: Complex 3D multi-parametric data using strong magnetic fields and radiofrequency pulses. Provides excellent soft tissue contrast without ionizing radiation. Uses strong magnetic fields and radio waves. Provides excellent soft tissue contrast customizable via multiple acquisition parameters. No ionizing radiation.

Ultrasound: Real-time dynamic imaging using sound wave reflections. Widely used for abdominal, pelvic, cardiac, and fetal exams due to low cost and lack of radiation. Employs soundwave echoes. Real-time imaging enabling visualization of moving structures. Portable and inexpensive but operator skill affects quality.