Impact of Original and Artificially Improved Artificial Intelligence-based Computer-aided Diagnosis on Breast US Interpretation.

OBJECTIVE: For breast US interpretation, to assess impact of computer-aided diagnosis (CADx) in original mode or with improved sensitivity or specificity. METHODS: In this IRB approved protocol, orthogonal-paired US images of 319 lesions identified on screening, including 88 (27.6%) cancers (median 7 mm, range 1-34 mm), were reviewed by 9 breast imaging radiologists. Each observer provided BI-RADS assessments (2, 3, 4A, 4B, 4C, 5) before and after CADx in a mode-balanced design: mode 1, original CADx (outputs benign, probably benign, suspicious, or malignant); mode 2, artificially-high-sensitivity CADx (benign or malignant); and mode 3, artificially-high-specificity CADx (benign or malignant). Area under the receiver operating characteristic curve (AUC) was estimated under each modality and for standalone CADx outputs. Multi-reader analysis accounted for inter-reader variability and correlation between same-lesion assessments. RESULTS: AUC of standalone CADx was 0.77 (95% CI: 0.72-0.83). For mode 1, average reader AUC was 0.82 (range 0.76-0.84) without CADx and not significantly changed with CADx. In high-sensitivity mode, all observers' AUCs increased: average AUC 0.83 (range 0.78-0.86) before CADx increased to 0.88 (range 0.84-0.90), P < 0.001. In high-specificity mode, all observers' AUCs increased: average AUC 0.82 (range 0.76-0.84) before CADx increased to 0.89 (range 0.87-0.92), P < 0.0001. Radiologists responded more frequently to malignant CADx cues in high-specificity mode (42.7% vs 23.2% mode 1, and 27.0% mode 2, P = 0.008). CONCLUSION: Original CADx did not substantially impact radiologists' interpretations. Radiologists showed improved performance and were more responsive when CADx produced fewer false-positive malignant cues.

Training Radiologists to Interpret Contrast-enhanced Mammography: Toward a Standardized Lexicon.

OBJECTIVE: Using terms adapted from the BI-RADS Mammography and MRI lexicons, we trained radiologists to interpret contrast-enhanced mammography (CEM) and assessed reliability of their description and assessment. METHODS: A 60-minute presentation on CEM and terminology was reviewed independently by 21 breast imaging radiologist observers. For 21 CEM exams with 31 marked findings, observers recorded background parenchymal enhancement (BPE) (minimal, mild, moderate, marked), lesion type (oval/round or irregular mass, or non-mass enhancement), intensity of enhancement (none, weak, medium, strong), enhancement quality (none, homogeneous, heterogeneous, rim), and BI-RADS assessment category (2, 3, 4A, 4B, 4C, 5). "Expert" consensus of 3 other radiologists experienced in CEM was developed. Kappa statistic was used to assess agreement between radiologists and expert consensus, and between radiologists themselves, on imaging feature categories and final assessments. Reproducibility of specific feature descriptors was assessed as fraction of consensus-concordant responses. RESULTS: Radiologists demonstrated moderate agreement for BPE, (mean kappa, 0.43; range, 0.05-0.69), and lowest reproducibility for "minimal." Agreement was substantial for lesion type (mean kappa, 0.70; range, 0.47-0.93), moderate for intensity of enhancement (mean kappa, 0.57; range, 0.44-0.76), and moderate for enhancement quality (mean kappa, 0.59; range, 0.20-0.78). Agreement on final assessment was fair (mean kappa, 0.26; range, 0.09-0.44), with BI-RADS category 3 the least reproducible. Decision to biopsy (BI-RADS 2-3 vs 4-5) showed moderate agreement with consensus (mean kappa, 0.54; range, -0.06-0.87). CONCLUSION: With minimal training, agreement for description of CEM findings by breast imaging radiologists was comparable to other BI-RADS lexicons.