Breast nodule classification with two-dimensional ultrasound using Mask-RCNN ensemble aggregation.

PURPOSE: The purpose of this study was to create a deep learning algorithm to infer the benign or malignant nature of breast nodules using two-dimensional B-mode ultrasound data initially marked as BI-RADS 3 and 4. MATERIALS AND METHODS: An ensemble of mask region-based convolutional neural networks (Mask-RCNN) combining nodule segmentation and classification were trained to explicitly localize the nodule and generate a probability of the nodule to be malignant on two-dimensional B-mode ultrasound. These probabilities were aggregated at test time to produce final results. Resulting inferences were assessed using area under the curve (AUC). RESULTS: A total of 460 ultrasound images of breast nodules classified as BI-RADS 3 or 4 were included. There were 295 benign and 165 malignant breast nodules used for training and validation, and another 137 breast nodules images used for testing. As a part of the challenge, the distribution of benign and malignant breast nodules in the test database remained unknown. The obtained AUC was 0.69 (95% CI: 0.57-0.82) on the training set and 0.67 on the test set. CONCLUSION: The proposed deep learning solution helps classify benign and malignant breast nodules based solely on two-dimensional ultrasound images initially marked as BIRADS 3 and 4.

Three artificial intelligence data challenges based on CT and ultrasound.

PURPOSE: The 2020 edition of these Data Challenges was organized by the French Society of Radiology (SFR), from September 28 to September 30, 2020. The goals were to propose innovative artificial intelligence solutions for the current relevant problems in radiology and to build a large database of multimodal medical images of ultrasound and computed tomography (CT) on these subjects from several French radiology centers. MATERIALS AND METHODS: This year the attempt was to create data challenge objectives in line with the clinical routine of radiologists, with less preprocessing of data and annotation, leaving a large part of the preprocessing task to the participating teams. The objectives were proposed by the different organizations depending on their core areas of expertise. A dedicated platform was used to upload the medical image data, to automatically anonymize the uploaded data. RESULTS: Three challenges were proposed including classification of benign or malignant breast nodules on ultrasound examinations, detection and contouring of pathological neck lymph nodes from cervical CT examinations and classification of calcium score on coronary calcifications from thoracic CT examinations. A total of 2076 medical examinations were included in the database for the three challenges, in three months, by 18 different centers, of which 12% were excluded. The 39 participants were divided into six multidisciplinary teams among which the coronary calcification score challenge was solved with a concordance index > 95%, and the other two with scores of 67% (breast nodule classification) and 63% (neck lymph node calcifications).

Phase-contrast MRI evaluation of haemodynamic changes induces by a coeliac axis stenosis in the gastroduodenal artery.

OBJECTIVE: To evaluate the correlation between the gastroduodenal artery (GDA) haemodynamic changes and the degree of coeliac axis (CA) stenosis using phase-contrast MRI. METHODS: The study was institutional review board approved, and written informed consent was obtained from patients included prospectively. A two-dimensional phase-contrast MRI was performed in 23 patients scheduled for a potential complex supramesocolic surgery, in a plane perpendicular to the GDA, during inspiration and expiration. The peak and mean velocities (Vp and Vm), mean flow rate (Qm) and flow direction at inspiration and at expiration have been correlated with the degree of CA stenosis evaluated by CT. RESULTS: 13 of 23 patients presented CA stenosis due to the median arcuate ligament (34-80% of stenosis), 4 of them had associated atheromatous calcifications. Vp, Vm and Qm of GDA presented a significant and linear relationship with the degree of CA stenosis, at inspiration as well as at expiration (r > 0.74, p < 0.001). Velocities and flow rates presented a different pattern depending on the degree of CA stenosis: <34% (n = 10), flow direction remained physiological with low velocities and flow rates; >60% of stenosis (n = 5), a reverse flow direction with increased velocities and flow rates were observed; variable patterns between 34% and 60%. CONCLUSION: Phase-contrast MRI permits the evaluation of haemodynamic changes in GDA induced by CA stenosis, including median arcuate ligament compression, and could be of great interest in therapeutic decision making in supramesocolic surgery, such as liver transplantation or duodenopancreatectomy, by detecting haemodynamically significant stenoses. Advances in knowledge: Physiological phase-contrast MRI detects haemodynamically significant stenoses of the CA by evaluating haemodynamic parameters in the GDA. Physiological phase-contrast MRI demonstrates that flow in the GDA can be a direct marker of the real haemodynamic impact of a CA stenosis on the hepatic vascularization and could thus participate in the pre-surgical work-up of duodenopancreatectomy or hepatic graft, whereas existing pre-operative imaging are only morphological.

Evaluation of recruited lung volume at inspiratory plateau pressure with PEEP using bedside digital chest X-ray in patients with acute lung injury/ARDS.

BACKGROUND: We wanted to assess whether there was a significant relationship between recruited lung volume (V(rec)) and change in density on digital processed chest x-ray measured at 2 different levels of inspiratory plateau pressure corresponding to 2 PEEP levels in patients with acute lung injury or ARDS. METHODS: In 14 subjects, PEEP 5 cm H2O and 15 cm H2O were prospectively applied in a random order for 10 min. At the end of each period, chest x-ray was taken using a digital portable device, and a pressure-volume curve of the respiratory system was performed. We also assessed P(aO2), and the static and the dynamic (C(dyn,rs)) compliance of the respiratory system. Change in end-expiratory lung volume between tidal breath and relaxation volume of the respiratory system was determined. Radiological attenuation was measured on chest x-rays in 4 regions of interest in the right lung, and in 3 regions of interest in the left lung, drawn in posterior intercostal spaces from top to bottom, by using dedicated software. The ratio of lung density in each region between PEEP 15 and PEEP 5 (rP15/P5) and their arithmetic mean (µP15/P5) were computed. V(rec) was determined from the pressure-volume curves. RESULTS: The median value of rP15/P5 in the 98 lung levels was 0.91 (0.80-1.01), which was significantly different from 1 (P < .001). The values of rP15/P5 were not significantly different between the lung levels. The median values of V(rec) and µP15/P5 were 288 (173-402) mL and 0.90 (0.80-0.97), respectively. There was a significant negative correlation between V(rec) and µP15/P5 (R = -0.77, P = .01). The reduction in µP15/P5 tended to correlate with the increase in C(dyn,rs) (R = -0.49, P = .077) or in P(aO2) (R = -0.53, P = .05) between PEEP 15 cm H2O and PEEP 5 cm H2O. CONCLUSIONS: Digital chest x-ray done at the bedside in acute lung injury/ARDS subjects was able to detect a reduction in density between PEEP 5 cm H2O and PEEP 15 cm H2O, which correlated with V(rec).