AI Tool Cuts Unnecessary Breast Biopsies, Saving Millions

The tool, referred to as iBRISK (intelligent-augmented breast cancer risk calculator), had been previously developed by scientists at Houston Methodist Hospital for a more accurate assessment of women's potential risk of developing breast cancer. The creation of this tool involved the application of deep learning techniques to clinical risk factors and descriptors from mammograms of 9,700 individuals at their institution. The model's validation was performed on an additional 1,000-plus patients.

In a recent investigation, the iBRISK tool underwent further testing using an independent retrospective dataset of breast images from an additional 4,209 patients screened across three Texas institutions from 2006 to 2016. This model was designed to specifically evaluate the likelihood of malignancy in Breast Imaging Reporting and Data System (BI-RADS) category 4 lesions.

The lead author, Chika F. Ezeana, along with colleagues from the Houston Methodist Neal Cancer Center, reported promising findings from the study. The AI tool demonstrated the potential to result in substantial cost savings by reducing unnecessary examinations.

According to the authors, "The iBRISK calculator can help physicians, primarily radiologists, categorize patients into low-probability of malignancy (POM) groups, thereby avoiding unnecessary biopsies of benign lesions. Conversely, high-risk groups can be treated as BI-RADS 5 patients." They also emphasized the importance of a refined stratification system that incorporates crucial patient characteristics alongside abnormal imaging features to enhance the accuracy of POM estimation. This improvement could guide the management of mammographic findings, minimize over-biopsy rates, reduce costs, and alleviate patient emotional distress.

The accuracy of the iBRISK model was calculated at approximately 89.5%, with an area under the receiver operating characteristic curve of 0.93. Specificity reached 81%, and out of the "low" POM group of 1,228 individuals, only two had malignant lesions. Conversely, the malignancy rate for the "high" POM group was 85.9%. The IBRISK score, serving as a continuous predictor of malignancy, achieved an area under the receiver operating characteristic curve of 0.97.

Ezeana and colleagues estimated that iBRISK could potentially save hundreds of millions annually for a single institution. This estimation was based on Medicare biopsy reimbursement rates and average costs for various types of tests. The researchers believe that implementing iBRISK could lead to cost savings of around $420 million across a sample of 390,000 women by avoiding unnecessary biopsies.

The authors concluded that their study highlighted the effectiveness of iBRISK in risk stratification for BI-RADS 4 lesions and in reducing unnecessary biopsies. They plan to publish the iBRISK calculator as an open-access online interface accessible to health systems and centers worldwide. Future research aims to enhance the model by incorporating more detailed data and addressing other BI-RADS categories.

---

By leveraging cutting-edge AI and IoT technology, VinBrain has successfully has been deployed in over 100 hospitals in Vietnam, Myanmar, New Zealand, India and the U.S. Our platforms are being used by thousands of doctors daily, leading to improved outcomes for more than 2 million patients and becoming The FIRST & ONLY AI Product for X-ray Diagnostics in Vietnam and Southeast Asia that meets US FDA standards. Check out our flagship products:
- DrAid™ CXR Screening
- DrAid™ CT Liver Cancer
- DrAid™ Enterprise Data
#DrAid #Digitalhealth #AIinHealthcare #AIinnovation #MedicalImaging #Radiology #Radiologist