AI drops DBT workloads for radiologists by 40% while also reducing recalls

A new artificial intelligence model can help spot normal breast screening exams and decrease the number of images that require radiologist interpretation by nearly 40%, according to data published Tuesday.

The tool was built using digital breast tomosynthesis exams taken from nearly 10,000 women and tested on more than 4,300. Compared to radiologists, the model yielded a nearly identical sensitivity while dropping recall rates by almost 25%.

DBT has been shown to improve cancer detection but also requires double the reading time as digital mammography scans. AI may prove useful as DBT continues to gain popularity among breast cancer screening programs around the world, researchers explained in Radiology.

“The artificial intelligence model was able to identify normal digital breast tomosynthesis screening examinations, which decreased the number of examinations that required radiologist interpretation in a simulated clinical workflow,” Yoel Shoshan, with the Department of Healthcare Informatics at IBM Research in Israel, and co-authors added. “Because 99.5% of screening examinations are cancer free, deploying such an AI system to optimize screening reads could be of substantial value.”

The retrospective investigation included 13,306 DBT exams performed across two healthcare networks between June 2013 and November 2018. All scans were split into a training group (3,948 women) validation set (1,661) and testing cohort (4,310).

Artificial intelligence detected 413 of 459 cancers (90%) compared to radiologists’ mark of 417 (90.8%) cancers detected. Using the tool to automatically filter out cases would drop a radiologists’ workload by 39.6%, the authors noted.

At the same time, the model did miss four cancer cases detected by radiologists, which is always a cause for concern, Liane E. Philpotts, MD, with the Department of Radiology and Biomedical Imaging at Yale, wrote in an editorial published alongside the study.

Philpotts brought up a number of important questions related to those misses, wondering whether radiologists could ever trust AI diagnoses knowing the system wasn’t 100% accurate.

“In reality, radiologists will likely still have to fully evaluate most DBT scans to detect those few subtle cancers, despite what the AI tells us,” the Yale imaging expert wrote.

 


Related Breast Imaging Technology Content:

DBT plus synthesized mammography drops patients’ dosage without losing imaging quality

3D mammography approaching 50% of breast imaging systems in the US

Ultrasound alone detects 92% of breast cancers recalled on DBT exams

Breast MRI screening cuts cancer mortality rates in half for women with lesser-known gene mutations

DBT spot compression views increase reader accuracy

Digital breast tomosynthesis outperforms DM at detecting malignancy in developing asymmetries

Less experienced radiologists are more susceptible to fatigue when reading DBT exams

Stand-alone AI can reduce radiologists’ screening mammography workloads by 90%

Imaging center has credentials revoked following ‘severe’ issues with mammography screening

Q&A: Should COVID vaccinated patients delay getting breast imaging — new study says no 

""

Matt joined Chicago’s TriMed team in 2018 covering all areas of health imaging after two years reporting on the hospital field. He holds a bachelor’s in English from UIC, and enjoys a good cup of coffee and an interesting documentary.

Around the web

The two companies aim to improve patient access to high-quality MRI scans by combining their artificial intelligence capabilities.

Positron, a New York-based nuclear imaging company, will now provide Upbeat Cardiology Solutions with advanced PET/CT systems and services. 

The nuclear imaging isotope shortage of molybdenum-99 may be over now that the sidelined reactor is restarting. ASNC's president says PET and new SPECT technologies helped cardiac imaging labs better weather the storm.