AsymMirai: Interpretable Mammography-based Deep Learning Model for 1-5-year Breast Cancer Risk Prediction.

Background Mirai, a state-of-the-art deep learning-based algorithm for predicting short-term breast cancer risk, outperforms standard clinical risk models. However, Mirai is a black box, risking overreliance on the algorithm and incorrect diagnoses. Purpose To identify whether bilateral dissimilarity underpins Mirai's reasoning process; create a simplified, intelligible model, AsymMirai, using bilateral dissimilarity; and determine if AsymMirai may approximate Mirai's performance in 1-5-year breast cancer risk prediction. Materials and Methods This retrospective study involved mammograms obtained from patients in the EMory BrEast imaging Dataset, known as EMBED, from January 2013 to December 2020. To approximate 1-5-year breast cancer risk predictions from Mirai, another deep learning-based model, AsymMirai, was built with an interpretable module: local bilateral dissimilarity (localized differences between left and right breast tissue). Pearson correlation coefficients were computed between the risk scores of Mirai and those of AsymMirai. Subgroup analysis was performed in patients for whom AsymMirai's year-over-year reasoning was consistent. AsymMirai and Mirai risk scores were compared using the area under the receiver operating characteristic curve (AUC), and 95% CIs were calculated using the DeLong method. Results Screening mammograms (n = 210 067) from 81 824 patients (mean age, 59.4 years ± 11.4 [SD]) were included in the study. Deep learning-extracted bilateral dissimilarity produced similar risk scores to those of Mirai (1-year risk prediction, r = 0.6832; 4-5-year prediction, r = 0.6988) and achieved similar performance as Mirai. For AsymMirai, the 1-year breast cancer risk AUC was 0.79 (95% CI: 0.73, 0.85) (Mirai, 0.84; 95% CI: 0.79, 0.89; P = .002), and the 5-year risk AUC was 0.66 (95% CI: 0.63, 0.69) (Mirai, 0.71; 95% CI: 0.68, 0.74; P < .001). In a subgroup of 183 patients for whom AsymMirai repeatedly highlighted the same tissue over time, AsymMirai achieved a 3-year AUC of 0.92 (95% CI: 0.86, 0.97). Conclusion Localized bilateral dissimilarity, an imaging marker for breast cancer risk, approximated the predictive power of Mirai and was a key to Mirai's reasoning. © RSNA, 2024 Supplemental material is available for this article See also the editorial by Freitas in this issue.

The Rashomon Importance Distribution: Getting RID of Unstable, Single Model-based Variable Importance.

Quantifying variable importance is essential for answering high-stakes questions in fields like genetics, public policy, and medicine. Current methods generally calculate variable importance for a given model trained on a given dataset. However, for a given dataset, there may be many models that explain the target outcome equally well; without accounting for all possible explanations, different researchers may arrive at many conflicting yet equally valid conclusions given the same data. Additionally, even when accounting for all possible explanations for a given dataset, these insights may not generalize because not all good explanations are stable across reasonable data perturbations. We propose a new variable importance framework that quantifies the importance of a variable across the set of all good models and is stable across the data distribution. Our framework is extremely flexible and can be integrated with most existing model classes and global variable importance metrics. We demonstrate through experiments that our framework recovers variable importance rankings for complex simulation setups where other methods fail. Further, we show that our framework accurately estimates the true importance of a variable for the underlying data distribution. We provide theoretical guarantees on the consistency and finite sample error rates for our estimator. Finally, we demonstrate its utility with a real-world case study exploring which genes are important for predicting HIV load in persons with HIV, highlighting an important gene that has not previously been studied in connection with HIV. Code is available at https://github.com/jdonnelly36/Rashomon_Importance_Distribution.

Closure of tubular patent ductus arteriosus in infants with the Amplatzer Vascular Plug II.

OBJECTIVES: We used the Amplatzer Vascular Plug II to close tubular patent ductus arteriosus (DA) in infants. BACKGROUND: Despite advancements in device design, catheter-based therapy for the DA of tubular morphology has been problematic. Likewise, the currently available devices are not designed to close DAs in small, often premature infants as the size of the delivery systems can be prohibitive and the devices obstructive to aortic or pulmonary artery flow. METHODS: We report our experience using the second-generation Amplatzer Vascular Plug (AVP II) in 10 patients with sizeable, tubular DAs, seven of whom were less than or equal to 4.0 kg. RESULTS: Complete closure was attained in all patients, with one minor complication. In four small infants, the device was delivered without arterial access under echocardiographic guidance. CONCLUSION: It is our belief that the AVP II device can be a useful embolization device for DAs in this difficult patient population.

Ductal origin of the pulmonary artery in isolation: a case series.

We present three cases of ductal origin of a pulmonary artery in the absence of associated structural heart disease. No geographic or genetic explanation for the temporal occurrence of these cases could be identified. This series illustrates the importance of angiography in achieving an accurate diagnosis as well as appropriate surgical planning. An understanding of the embryologic origin of this lesion supports the strategic approach at catheterization as well as the surgical choice of interposition graft versus direct anastomosis. We report our experience and highlight diagnostic and surgical approaches to such patients based on the underlying embryology of this congenital cardiac anomaly.

Use of Amplatzer occlusion devices to occlude Fontan baffle leaks during fenestration closure procedures.

The authors describe the use of Amplatzer occlusion devices to successfully occlude Fontan baffle leaks in three patients. These additional right to left shunts were recognized at the time of elective transcatheter closure of surgically created fenestrations. A common site of baffle leak is described at the base of the right atrial appendage. This suture line serves to exclude the right atrial appendage from Fontan baffle flow and lies within the corrugated surface of the pectinate muscles, making it particularly vunerable to baffle leaks over time.

Childhood cardiac function after prenatal diagnosis of intracardiac echogenic foci.

OBJECTIVE: To determine whether prenatally diagnosed intracardiac echogenic foci are associated with childhood cardiac dysfunction and persistence. METHODS: Children in whom intracardiac echogenic foci were shown on prenatal sonography at 1 perinatal center underwent echocardiography at ages 2 to 7 years. A single pediatric cardiologist, blinded to the prenatal sonographic intracardiac echogenic focus locations, assessed cardiac function by measuring the left ventricular shortening fraction and myocardial performance index. The presence of tricuspid and mitral valve regurgitation was also sought. The secondary outcome was intracardiac echogenic focus persistence. RESULTS: Twenty-five children, 14 (56%) male and 11 (44%) female, were examined at a mean age +/- SD of 3.0 +/- 1.0 years. Prenatally, 18 children (72%) had left ventricular intracardiac echogenic foci, and 7 (28%) had right ventricular intracardiac echogenic foci. The left ventricular shortening fraction was normal in all children. The overall mean left ventricular myocardial performance index (reference value, 0.36 +/- 0.06), was normal for both children with left ventricular intracardiac echogenic foci (0.36 +/- 0.06) and those with right ventricular intracardiac echogenic foci (0.36 +/- 0.04). Two children with left ventricular intracardiac echogenic foci had an isolated left ventricular myocardial performance index of greater than 2.5 SD above the mean. Trace tricuspid valve regurgitation and mitral valve regurgitation were noted in 13 (52%) and 2 (8%) of the children, respectively, similar to the general population. Left ventricular intracardiac echogenic foci persisted in 16 children (89%), whereas right ventricular intracardiac echogenic foci persisted in 2 (29%) (P = .007). CONCLUSIONS: Prenatally diagnosed intracardiac echogenic foci are often persistent but not associated with childhood myocardial dysfunction.

Revascularization of a thrombosed aortopulmonary shunt with the use of the AngioJet thrombectomy system.

We report the use of the AngioJet F140 rheolytic catheter to recannalize an acutely thrombosed aortopulmonary shunt in a 21-year-old female with palliated, complex congenital heart disease. After extracting the thrombus that filled the entire length of the shunt, three stents were placed at sites where the lumen was compromised by distortion or thrombus. Unobstructed flow was restored to the left pulmonary artery that persisted at 1-year follow-up.

Repair of hypoplastic left heart syndrome of a 4.25-kg Jehovah's witness.

The care of patients who refuse homologous transfusions has challenged cardiac surgery teams to refine blood conservation techniques and question standard transfusion practices. We cared for a newborn child with hypoplastic left heart syndrome (HLHS) whose parents refused to give consent to care for the child that involved the transfusion of homologous blood. A Norwood Stage I procedure was planned with the understanding that transfusions would be avoided, if possible. A court order was obtained that specified the conditions under which the attending physicians would transfuse the newborn. The birth weight of the patient was 4.25 kg. A low prime cardiopulmonary bypass (CPB) circuit and aggressive blood conservation techniques that included modified ultrafiltration (MUF) allowed the completion of the repair and CPB portion of the operation without the use of blood. The lowest hematocrit during CPB was 20%. After an unsuccessful attempt to separate from CPB, blood was transfused. Recovery was consistent for HLHS patients following Norwood Stage I. However, at 1 month postoperatively, the patient did require a shunt reduction for pulmonary overcirculation. Norwood Stage II repair was completed at age 4 months without donor blood. The key to a successful outcome is a well-thoughtout plan by the surgeon, anesthesiologist and perfusionist. This plan should include careful monitoring of the patient's oxygenation and cardiovascular status.