Background

The commonly used diagnostic method, ultrasonography (USG), is considered subjective and largely dependent on the operator’s experience. For several years, research has been emerging indicating the increasing role of artificial intelligence (AI) as an additional tool in ultrasound examinations, aimed at supporting clinicians in obtaining test results with the highest possible sensitivity and precision.

Objectives

USG is the method of choice for imaging thyroid nodules, thus selecting them for fine-needle biopsy. To standardize test results and reduce the risk of the operator’s experience influencing the results, a computer-aided image analysis system was introduced.

Material and methods

A USG examination was performed in three patients, with the AI-assisted ultrasound examination results in a report containing the relevant components.

Results

The dimensions of the nodules determined in the transverse plane (width and depth) are comparable. Considering that the examination description proposed by AI may be incorrect, it can be manually modified by the operator and then confirmed. Once approved, the description is imported into the report, along with the three dimensions of the nodule and the calculated volume, and is ready for printing.

Conclusions

1. Many studies compare AI performance directly with radiologists’ performance, often showing AI as a valuable addition, especially for less experienced clinicians. 2. AI systems are increasingly integrated into clinical workflows, with some embedded in ultrasound equipment and others being developed as web-based tools or decision support systems. 3. Interpretability and explainability remain crucial for clinical acceptance, with methods such as Grad-CAM and interpretable multi-attribute networks being developed.