Using deep learning to safely exclude lesions with only ultrafast breast MRI to shorten acquisition and reading time.

OBJECTIVES: To investigate the feasibility of automatically identifying normal scans in ultrafast breast MRI with artificial intelligence (AI) to increase efficiency and reduce workload. METHODS: In this retrospective analysis, 837 breast MRI examinations performed on 438 women from April 2016 to October 2019 were included. The left and right breasts in each examination were labelled normal (without suspicious lesions) or abnormal (with suspicious lesions) based on final interpretation. Maximum intensity projection (MIP) images of each breast were then used to train a deep learning model. A high sensitivity threshold was calculated based on the detection trade - off (DET) curve on the validation set. The performance of the model was evaluated by receiver operating characteristic analysis of the independent test set. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with the high sensitivity threshold were calculated. RESULTS: The independent test set consisted of 178 examinations of 149 patients (mean age, 44 years ± 14 [standard deviation]). The trained model achieved an AUC of 0.81 (95% CI: 0.75-0.88) on the independent test set. Applying a threshold of 0.25 yielded a sensitivity of 98% (95% CI: 90%; 100%), an NPV of 98% (95% CI: 89%; 100%), a workload reduction of 15.7%, and a scan time reduction of 16.6%. CONCLUSION: This deep learning model has a high potential to help identify normal scans in ultrafast breast MRI and thereby reduce radiologists' workload and scan time. KEY POINTS: • Deep learning in TWIST may eliminate the necessity of additional sequences for identifying normal breasts during MRI screening. • Workload and scanning time reductions of 15.7% and 16.6%, respectively, could be achieved with the cost of 1 (1 of 55) false negative prediction.

Image-guided point-shear-wave elastography: a valid and reliable technique for liver fibrotic staging.

PURPOSE: Despite progressive implementation of image-guided point-shear wave elastography (pSWE) in guidelines as an alternative to transient elastography for the staging of fibrotic liver disease, pSWE is not widely adopted in clinical workflow. More information on reliability and validity of pSWE systems is needed. Therefore, we performed a phantom study to evaluate the validity and reliability of pSWE with ultrasound systems. METHODS: Validity and reliability of pSWE measurements from three ultrasound systems were evaluated. Measurements were performed on an elasticity phantom with reference elasticities of 7 ± 1 (low) (median ± interquartile range (IQR)), 14 ± 2 (medium) and 26 ± 3 (high) kPa. Measurements were repeated in tenfold for each reference at 2, 3 and 4 cm depth. Results were considered valid when median elasticity ± IQR was between the uncertainty limits (IQR) for each reference elasticity value and reliable when IQR/median < 0.30. RESULTS: pSWE with the systems provided valid results for all reference elasticities and focal depths, except for overestimation of high reference elasticity at 2 and 4 cm depth for one system (41.5 ± 4.3 and 39.0 ± 1.2 kPa, respectively). Measurements were reliable with a maximum IQR/median of 0.13, well below the guideline of IQR/median < 0.30. DISCUSSION: The results support the use of pSWE as an alternative to invasive or non-image guided noninvasive techniques for liver fibrotic staging. CONCLUSIONS: pSWE with ultrasound systems from different vendors is valid and reliable and can therefore be implemented to optimize clinical workflow by performing imaging and elastography simultaneously.

Dose Reduction in Coronary Artery Calcium Scoring Using Mono-Energetic Images from Reduced Tube Voltage Dual-Source Photon-Counting CT Data: A Dynamic Phantom Study.

In order to assess coronary artery calcium (CAC) quantification reproducibility for photon-counting computed tomography (PCCT) at reduced tube potential, an anthropomorphic thorax phantom with low-, medium-, and high-density CAC inserts was scanned with PCCT (NAEOTOM Alpha, Siemens Healthineers) at two heart rates: 0 and 60-75 beats per minute (bpm). Five imaging protocols were used: 120 kVp standard dose (IQ level 16, reference), 90 kVp at standard (IQ level 16), 75% and 45% dose and tin-filtered 100 kVp at standard dose (IQ level 16). Each scan was repeated five times. Images were reconstructed using monoE reconstruction at 70 keV. For each heart rate, CAC values, quantified as Agatston scores, were compared with the reference, whereby deviations >10% were deemed clinically relevant. Reference protocol radiation dose (as volumetric CT dose index) was 4.06 mGy. Radiation dose was reduced by 27%, 44%, 67%, and 46% for the 90 kVp standard dose, 90 kVp 75% dose, 90 kVp 45% dose, and Sn100 standard dose protocol, respectively. For the low-density CAC, all reduced tube current protocols resulted in clinically relevant differences with the reference. For the medium- and high-density CAC, the implemented 90 kVp protocols and heart rates revealed no clinically relevant differences in Agatston score based on 95% confidence intervals. In conclusion, PCCT allows for reproducible Agatston scores at a reduced tube voltage of 90 kVp with radiation dose reductions up to 67% for medium- and high-density CAC.