[Machine learning in radiology : Terminology from individual timepoint to trajectory]

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Machine learning (ML) algorithms are increasingly vital in radiology.
• They are applied across various imaging modalities, including CT, MRI, and ultrasound.
• Each modality presents unique challenges in standardization and data variability.

Purpose of the Study:

• To explore the role and impact of machine learning (ML) algorithms in modern radiology.
• To highlight the capabilities of ML in analyzing medical images for diagnosis and prediction.
• To discuss the current state and future implications of ML in radiological practice.

Main Methods:

• Utilizing ML algorithms for automatic detection and segmentation of diagnostic markers.
• Employing ML for quantifying disease progression and treatment response.
• Training prediction models using longitudinal data for personalized patient outcomes.

Main Results:

• ML algorithms demonstrate acceptable performance in lesion detection and segmentation.
• The accuracy of ML-based prediction models is continuously improving.
• ML solutions in radiology are progressing, with many currently in the validation phase.

Conclusions:

• The successful integration of ML into radiology hinges on rigorous algorithm validation.
• The development of representative and diverse datasets is essential for training and validating ML models.
• ML holds significant potential to support clinical decisions and advance radiological practice.