Using Artificial Intelligence to Support Efficient Same-Day Diagnostic Imaging in Breast Cancer Screening

Use of artificial intelligence algorithms that enable immediate interpretation of screening mammograms and, if needed, same-day diagnostic imaging has the potential to vastly shorten the time from an abnormal screening mammogram to diagnostic workup, resolve false positives in a timelier fashion, and reduce the anxiety incurred in patients by long wait times for a diagnostic evaluation.

Project Details -
Ongoing

Grant Number

R21 HS029257
Funding Mechanism(s)

Using Innovative Digital Healthcare Solutions to Improve Quality at the Point of Care (R21/R33 - Clinical Trial Optional)
AHRQ Funded Amount

$1,003,772
Principal Investigator(s)

Hsu, William

Hoyt, Anne C
Organization

University of California, Los Angeles
Location

Los Angeles

California
Project Dates

09/30/2022 - 09/29/2027
Technology

Artificial Intelligence

Machine Learning
Care Setting

Ambulatory Setting
Medical Condition

Breast Cancer

Cancer
Population

Women
Type of Care

Primary Care

Specialty Care
Health Care Theme

Access to Care

Patient-Centered Care

Breast cancer is the most common solid cancer and the second leading cause of cancer death among US women. While mammography screenings have led to a 25-40 percent reduction in breast cancer mortality, associated harms include false-positive results and the delays and anxiety incurred after an abnormal screen waiting for additional diagnostic imaging. Ten percent of all screening exams require further diagnostic workup; 95 percent of these are ultimately false positives. In addition to further workups resulting in benign biopsies and overdiagnosis, the amount of anxiety incurred to patients during this time waiting for a final diagnosis is significant.

This research will develop, implement, and evaluate an artificial intelligence (AI)-enabled workflow that would allow for same-day diagnostic imaging following screening mammography in high-volume breast cancer screening clinics. AI algorithms would facilitate the immediate interpretation of screening mammograms, normally done after a patient leaves the clinic, and allow those needing further imagining to be scheduled for a diagnostic exam on the same day.

The specific aims of the research are as follows:

Validate and integrate an AI algorithm to triage screening mammograms.
Design and assess an AI-enabled workflow for same-day diagnostic exams.
Implement and evaluate the impacts of an AI-enabled same-day diagnostic imaging paradigm.

The AI algorithms will be validated using existing data from the breast screening population at the researchers’ institution. Using insights from the validation step, the researchers will evaluate the current state of care, identify barriers to implementing such a program, and make changes to current workflows to mitigate any potential impacts of the AI intervention, ensuring patients receive the same level of care as current radiologist standards. Potential impacts include the number of patients who can be seen at the clinic, the workload of the breast imaging clinic staff, and the number of same-day diagnostic exams that would need to be accommodated. The evaluation phase of the research will include a pilot at a single imaging center with the eventual goal of expanding the program to breast imaging centers across the health system.

The impact of this research is expected to be a reduction in callback rates for women with abnormal screening, an increase in patient satisfaction from same-day interpretation, and the elimination of the delay between screening and diagnostic workup. This research will also identify strategies for implementing and evaluating the impact of AI on high-volume clinical workflows such as breast screening.

Using discrete event simulation to design and assess an AI-aided workflow for same-day diagnostic testing of women undergoing breast screening.

Citation: Lin Y, Hoyt AC, Manuel VG, Inkelas M, Hsu W. Using discrete event simulation to design and assess an AI-aided workflow for same-day diagnostic testing of women undergoing breast screening. AMIA Jt Summits Transl Sci Proc. 2024 May 31;2024:314-323. PMID: 38827101; PMCID: PMC11141813.

Link: https://pubmed.ncbi.nlm.nih.gov/38827101/

Principal Investigator: Hsu, William, Hoyt, Anne C

Project Name: Using Artificial Intelligence to Support Efficient Same-Day Diagnostic Imaging in Breast Cancer Screening

Document Type: Conference Proceeding

Research Method: Descriptive

Technology: Artificial Intelligence, Machine Learning

Population: Women

Use of Artificial Intelligence to Support Same-Day Breast Cancer Diagnostic Testing

Theme:

Optimizing Care Delivery for Clinicians

Subtheme:

Using Real-Time Digital Healthcare Data to Improve Timely Treatment or Diagnosis

Use of an artificial intelligence algorithm that would allow for screening mammography interpretation and same-day diagnostic imaging has the potential to vastly shorten the time from an abnormal screening mammogram to diagnostic workup, resolving false positives in a timelier fashion, and reducing the anxiety incurred in patients by long wait times for diagnostic evaluation.

While breast cancer screening can save lives, false positives are common

Mammography screening in the United States has been an incredibly successful tool for identifying breast cancer and has led to a 25-40 percent reduction in breast cancer-related deaths. While most screenings reveal normal results, approximately 10 percent of all screenings require a followup diagnostic exam to confirm a cancer diagnosis. Of women recalled for additional diagnostic exams, roughly 95 percent do not receive a diagnosis of breast cancer.

While this is good news, breast cancer is scary. And the need to return for a diagnostic exam can induce substantial anxiety and distress in women as they wait for their diagnostic appointment and results.

Artificial intelligence will improve time to diagnostic exam

Dr. William Hsu, an informaticist, and Dr. Anne Hoyt, a breast radiologist, both from the University of California Los Angeles, wanted to find a way to speed up the process to reduce patient stress and anxiety. Breast radiologists typically review screening mammograms in batches. This can occur anywhere from the day of the mammogram to more than a week later, depending on workload and staff availability. Identifying abnormal screenings immediately and conducting diagnostic exams on the same day would alleviate patient stress and enable a transformative “one-stop-shop” paradigm for breast cancer screening and diagnosis.

To do this, the team is using artificial intelligence (AI) algorithms at several steps throughout their workflow to immediately interpret screening mammograms and set up same-day diagnostic imaging, if required. This would allow women to undergo their regularly scheduled mammogram and any necessary diagnostic exam on the same day.

“AI is a tool that can help us to be better radiologists, more efficient radiologists, and ultimately to benefit the patients by finding cancers more quickly and initiate treatment sooner.”- Dr. Anne Hoyt

Same-day diagnostic exams supported by AI expected to improve care for women

The team will validate various AI algorithms for assessing breast cancer risk and detecting breast cancers on screening mammograms. Using the information on the performance of the AI algorithms, they will investigate how the implementation of an AI-enabled same-day diagnostic exam workflow would impact the number of patients who can be seen at the clinic, the workload of the breast imaging clinic staff, and the number of same-day diagnostic exams that would need to be accommodated. These insights will inform how to integrate the algorithms into clinical workflow to increase efficiency, maximize patient throughput, and promptly communicate results and appropriate next steps to patients while performing at the same level as current radiologist standards. The team expects that imaging centers using AI to provide immediate interpretation of screening exams and same-day diagnostic exams, as needed, will have lower callback rates of women with abnormal screening, increased patient satisfaction from same-day interpretation, and shorter times between screening and diagnostic workup when compared to centers that use the current state of mammography review.