Ferrum Health partners with best-in-class AI providers offering solutions designed to reduce radiology workload and improve patient care.
TaiHao Case Study: AI Can Alleviate Breast Ultrasound Workload
Increasing efficiency and easing the workload are priorities for radiologists. According to the dataset released by NHS England and NHS Improvement, 41.1 million imaging tests were reported in England during the 12 months, November 2020 to October 2021. However, statistics shows the annual number of medical radiographers in the same period is approximately 37.3 thousand. Undoubtedly, the disparity between the two figures between imaging and radiographers makes AI diagnostic decision support more indispensable. From RSNA annual meeting in 2021, we could also realize that artificial intelligence (AI) can reduce false-positive findings and potentially eliminate up to 80% of breast ultrasound exams from the radiologist worklist.
Reader Case Study
The purpose of this case study was to compare the diagnostic performance and interpretation time of breast ultrasound examination between reading without and with an artificial intelligence system, BU-CAD TaiHao Medical, as a concurrent reading aid.
Methods and Evaluation Scenarios
Fourteen radiologists and two breast surgeons validated the reader study. Each reader reviewed 172 breast ultrasound cases in unaided and aided reading scenarios. Interpretations of any given case set with and without the AI system were separated by at least 5 weeks. Results were compared to the reference standard, and the area under the LROC curve, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for performance evaluations. The interpretation time was also compared between the unaided and aided scenarios.
Result – Reader LROC Analysis
With the AI system assistance, readers have a higher diagnostic performance with an increased average AUCLROC from 0.7582 to 0.8294, which is statistically significant, as shown in Figure
Result – Interpretation Time
The average interpretation time is significantly reduced by approximately 40% when readers are aided by the AI system, as shown in Table 2.
A 30-year-old patient had a physical examination after feeling unwell in her right breast. After her examination, radiologists reviewed the ultrasound images utilizing BU-CAD TaiHao Medical, an AI system. A suspicious soft tissue lesion in each ultrasound image was detected.
As shown in Figure 2, the software annotated the image by creating a bounding box surrounding the lesion, depicted the lesion contour with dots, and applied output diagnosis results such as Score of Lesion Characteristics (SLC), BI-RADS Category, and BI-RADS Descriptors.
The radiologists and breast surgeons improved their diagnostic performance in detecting and diagnosing breast lesions on breast ultrasound images with the assistance of BU-CAD, an AI system.
Given the large volume of cases performed daily, which is only increasing, it is essential to have a second set of eyes to assist radiologists. In this situation, an AI tool such as BU-CAD can improve the diagnosis performance, speed up diagnostics, and reduce the workload of radiologists.
Interested in deploying BU-CAD at your health facility?