AI outperforms humans in radiology reporting accuracy

A recent study from Harvard Medical School has revealed issues in how automated systems assess the accuracy of AI-generated radiology reports compared to human radiologists. To address this, two improved scoring systems, RadGraph F1 and RadCliQ, were introduced, which perform better in identifying clinical errors. These systems can boost the development and adoption of AI in radiology by ensuring more reliable assessments of AI model performance, improving patient care.