Impact of Artificial Intelligence on Women's Imaging: Cost-Benefit Analysis.

OBJECTIVE. The purpose of this article is to identify and discuss four areas in which artificial intelligence (AI) must excel to become clinically viable: performance, time, work flow, and cost. CONCLUSION. AI holds tremendous potential for transforming the practice of radiology, but certain metrics are needed to objectively quantify its impact. As patients, physicians, hospitals, and insurance companies look for value, AI must earn a role in medical imaging.

Screening Mammography Findings From One Standard Projection Only in the Era of Full-Field Digital Mammography and Digital Breast Tomosynthesis.

OBJECTIVE: The purpose of this article is to compare outcomes of findings seen on one view only from screening full-field digital mammography (FFDM) and FFDM plus digital breast tomosynthesis (DBT). MATERIALS AND METHODS: A retrospective review was performed of 103,070 consecutive screening mammograms obtained from February 2011 through June 2014 at two community-based breast centers. Recalled findings prospectively seen on one view only were analyzed. Pearson chi-square test and Fisher exact test were performed, and 95% CIs were determined. RESULTS: Mammograms were acquired using FFDM (n = 71,656) and FFDM-DBT (n = 31,414) during the study period, and 2213 FFDM (3.1%) and 433 FFDM-DBT (1.4%) mammograms were recalled for one-view-only findings (p < 0.0001). The final study population (1592 FFDM and 354 FFDM-DBT) was defined after 689 of these recalled FFDM and 92 of these recalled FFDM-DBT examinations were excluded for insufficient mammographic follow-up (< 24 months). Summation artifacts accounted for more one-view-only findings from FFDM (1067/1592; 67.0%) than FFDM-DBT (190/354; 53.7%) (p < 0.0001). In the FFDM cohort, 28 one-view-only findings proved malignant (24 invasive ductal carcinoma [IDC], one invasive lobular carcinoma [ILC], and three ductal carcinoma in situ [DCIS]). In the FFDM-DBT cohort, 14 one-view-only findings proved malignant (11 IDC, one ILC, and two DCIS). The positive predictive value (PPV) of a one-view-only finding was significantly lower for FFDM (1.8%) than for FFDM-DBT (4.0%) (p = 0.010). CONCLUSION: One-view-only findings occur with both FFDM and FFDM-DBT and remain an important but uncommon sign of malignancy. They are more frequent, are more likely to represent summation artifacts, and have a lower PPV with FFDM than with FFDM-DBT.

Digital Breast Tomosynthesis in the Diagnostic Setting: Indications and Clinical Applications.

Digital breast tomosynthesis (DBT) is an emerging technology used in diagnostic breast imaging to evaluate potential abnormalities. In DBT, the compressed breast tissue is imaged in a quasi-three-dimensional manner by performing a series of low-dose radiographic exposures and using the resultant projection image dataset to reconstruct cross-sectional in-plane images in standard mammographic views. Improved visualization of breast detail at diagnostic DBT allows improved characterization of findings, including normal structures and breast cancer. This technology reduces the summation of overlapping breast tissue, which can mimic breast cancer, and provides improved detail of noncalcified mammographic findings seen in breast cancer. It also assists in lesion localization and determining mammographic extent of disease in women with known or suspected breast cancer. The authors review the potential uses, benefits, and limitations of DBT in the diagnostic setting and discuss how radiologists can best use DBT to characterize lesions, localize potential abnormalities, and evaluate the extent of known or suspected breast cancer. The authors' experience shows that DBT can be implemented effectively in the diagnostic workflow to evaluate and localize potential lesions more efficiently. DBT may potentially replace conventional supplemental mammography at diagnostic workup and obviate ultrasonography in select cases.

Potential Impact of an Artificial Intelligence-based Mammography Triage Algorithm on Performance and Workload in a Population-based Screening Sample.

OBJECTIVE: To evaluate potential screening mammography performance and workload impact using a commercial artificial intelligence (AI)-based triage device in a population-based screening sample. METHODS: In this retrospective study, a sample of 2129 women who underwent screening mammograms were evaluated. The performance of a commercial AI-based triage device was compared with radiologists' reports, actual outcomes, and national benchmarks using commonly used mammography metrics. Up to 5 years of follow-up examination results were evaluated in cases to establish benignity. The algorithm sorted cases into groups of "suspicious" and "low suspicion." A theoretical workload reduction was calculated by subtracting cases triaged as "low suspicion" from the sample. RESULTS: At the default 93% sensitivity setting, there was significant improvement (P <.05) in the following triage simulation mean performance measures compared with actual outcome: 45.5% improvement in recall rate (13.4% to 7.3%; 95% CI, 6.2-8.3), 119% improvement in positive predictive value (PPV) 1 (5.3% to 11.6%; 95% CI, 9.96-13.4), 28.5% improvement in PPV2 (24.6% to 31.6%; 95% CI, 24.8-39.1), 20% improvement in sensitivity (83.3% to 100%; 95% CI, 100-100), and 7.2% improvement in specificity (87.2% to 93.5%; 95% CI, 92.4-94.5). A theoretical 62.5% workload reduction was possible. At the ultrahigh 99% sensitivity setting, a theoretical 27% workload reduction was possible. No cancers were missed by the algorithm at either sensitivity. CONCLUSION: Artificial intelligence-based triage in this simulation demonstrated potential for significant improvement in mammography performance and predicted substantial theoretical workload reduction without any missed cancers.

Metabolic Responses to Metformin in Inoperable Early-stage Non-Small Cell Lung Cancer Treated With Stereotactic Radiotherapy: Results of a Randomized Phase II Clinical Trial.

BACKGROUND: Metformin reduces glucose uptake in physiologic tissues and has been shown to affect non-small cell lung cancer (NSCLC) metabolism. We hypothesized that positron emission tomography (PET) scans could detect the impact of metformin on glucose uptake in NSCLC and we sought to test this hypothesis in a prospective clinical trial. MATERIALS AND METHODS: A single-blinded phase II clinical trial was performed with subjects randomized 6:1 to 3 to 4 weeks of metformin versus placebo for inoperable early-stage NSCLC. PET scans were performed at baseline, mid-treatment (after 2 wk study medication), and 6 months postradiation. The primary endpoint of the trial was tumor metabolic response to metformin by PERCIST before definitive radiation. Stereotactic body radiotherapy to 50 Gy in 4 fractions was used for peripheral tumors and 70 Gy in 10 fractions for central tumors. RESULTS: There were 14 subjects randomized to the metformin and 1 to placebo. Histologies were 60% adenocarcinoma, 33.3% squamous cell carcinoma, and 6.7% poorly differentiated carcinoma. At mid-treatment PET scan, 57% of subjects randomized to metformin met PERCIST criteria for metabolic response, of which 75% had progressive metabolic disease and 25% had partial metabolic response, whereas the placebo subject had stable metabolic disease. At 6 months, the metformin arm had 69% complete metabolic response, 23% partial metabolic response and 1 progressive metabolic disease, and the subject treated with placebo had a complete metabolic response. There were no CTCAE grade ≥3 toxicities. CONCLUSIONS: Despite low accrual, majority of subjects treated with metformin had metabolic responses by PERCIST criteria on PET imaging. Contrary to the effect of metformin on most physiologic tissues, most tumors had increased metabolic activity in response to metformin.

Preoperative localization of breast lesions: Current techniques.

Image-guided preoperative localization of breast lesions is a common procedure. This article describes several commercially available localization options-wire localization, radioactive seed localization, localization with a radiofrequency reflector, and magnetic seed localization-and outlines the advantages and disadvantages of each. This information may help radiologists initiate conversations at their facilities with surgeons, pathologists, and hospital administration as they seek to add value and provide patient-centered care.

Impact of Second-Opinion Interpretation of Breast Imaging Studies in Patients Not Currently Diagnosed With Breast Cancer.

PURPOSE: To study the impact of second-opinion interpretation of breast imaging studies submitted from outside facilities to a tertiary cancer center. MATERIALS AND METHODS: A retrospective database review was conducted of second-opinion interpretations rendered at our institution from January 1, 2010, to June 30, 2014, on studies from patients who did not have a concurrent breast cancer diagnosis. A total of 2,253 patients were included. RESULTS: In 800 of 2,253 patients (35.5%), the BI-RADS categories assigned at our institution and at outside facilities were discordant. Of 973 patients assigned BI-RADS category 4 or 5 at outside facilities, 278 (28.6%) were assigned BI-RADS category 1 to 3 (no biopsy necessary) at our institution. Of 923 patients assigned BI-RADS category 1 to 3 at outside facilities, 191 (20.7%) were assigned BI-RADS category 4 or 5 at our institution, and 189 of these had biopsies, which revealed 23 cancers, 15 high-risk lesions, and 151 benign lesions. One high-risk lesion at core biopsy was upgraded to invasive ductal carcinoma and ductal carcinoma in situ (DCIS) on excision, resulting in 24 cancers. Of these, 18 reflected true additional breast cancers detected as a result of second-opinion interpretation: 12 invasive carcinomas and 6 cases of DCIS. These results translate into a 9.4% (18/191) positive predictive value for the number of cancers diagnosed among all biopsies recommended and a 9.5% (18/189) positive predictive value for the number of cancers diagnosed among all biopsies recommended and actually performed. CONCLUSIONS: These findings demonstrate the positive clinical impact of second-opinion interpretation at a tertiary cancer center of outside-facility breast imaging studies in patients without a breast cancer diagnosis.

Middle cerebral artery aneurysms: a single-center series comparing endovascular and surgical treatment.

OBJECTIVE: The optimal treatment for middle cerebral artery (MCA) aneurysms is controversial. MCA aneurysms have been considered more conducive to surgical treatment. Recent technology has led to successful endovascular treatment of MCA aneurysms. The objective of this study was to analyze the outcomes of endovascular and surgical treatment of MCA aneurysms as experienced by a single tertiary center. METHODS: We retrospectively reviewed 90 MCA aneurysms in 84 patients treated from 2005 to 2010. They were separated into 2 groups: endovascular coiling, with 50 (59.5%) patients, and surgical clipping, with 34 (40.5%) patients. Outcome was based on complications, procedural morbidity and mortality, clinical and angiographic outcomes, and retreatment rates. Patients were further separated into ruptured and unruptured aneurysm groups. RESULTS: Ruptured aneurysms were 10 of 50 (20%) and 9 of 34 (26.5%) patients in the endovascular and surgical groups, respectively. Procedure-related complications were 16% and 0% for the endovascular and surgical groups (P = .01), respectively. Overall rate of complete or near-complete occlusion at angiographic follow-up was 86% and 95% for the endovascular and surgical groups (P = .16), respectively. Proportion of patients with modified Rankin scale of 3 to 6 at 6 months follow-up was 10% and 5.9% for the endovascular and surgical groups (P = .5), respectively. The mean angiographic follow-up was 9.02 months (range 0 to 5.2 years). Retreatment rates were 14% and 0% for the endovascular and surgical groups, respectively (P = .01). CONCLUSIONS: In this nonrandomized sample of 90 MCA aneurysms treated with endovascular coiling or neurosurgical clipping, we observed a similar clinical outcome based on the modified Rankin scale and angiographic occlusion. Complication and retreatment rates were higher but not significant for the endovascular group. Both treatment modalities are good alternatives and should be individualized based on aneurysm angioarchitecture and the patient's general conditions.