This thesis addresses two challenges in medical imaging: ensuring
complete coverage during breast ultrasound examinations and providing automated
evaluation of exercises in speech therapy. For breast ultrasound, an assistive
evaluation system based on electromagnetic tracking data was developed and
validated through a study of 120 examinations, revealing that even experienced
radiologists miss regions larger than 1 cm² in 3% of cases. The system enables
real-time visualization of examination coverage and provides quantitative metrics
for quality assessment. In speech therapy, a novel computer vision framework
(ASSISLT) was developed that combines deep learning-based tongue detection,
feature-based lip tracking, teeth detection, and brightness distribution analysis
for cheek movement evaluation. The system’s effectiveness was demonstrated
through controlled experiments with 20 participants, showing a 50% improvement
in exercise validity.