Automated Breast Density Assessment in MRI Using Deep Learning and Radiomics: Strategies for Reducing Inter-Observer Variability.

BACKGROUND: Accurate breast density evaluation allows for more precise risk estimation but suffers from high inter-observer variability. PURPOSE: To evaluate the feasibility of reducing inter-observer variability of breast density assessment through artificial intelligence (AI) assisted interpretation. STUDY TYPE: Retrospective. POPULATION: Six hundred and twenty-one patients without breast prosthesis or reconstructions were randomly divided into training (N = 377), validation (N = 98), and independent test (N = 146) datasets. FIELD STRENGTH/SEQUENCE: 1.5 T and 3.0 T; T1-weighted spectral attenuated inversion recovery. ASSESSMENT: Five radiologists independently assessed each scan in the independent test set to establish the inter-observer variability baseline and to reach a reference standard. Deep learning and three radiomics models were developed for three classification tasks: (i) four Breast Imaging-Reporting and Data System (BI-RADS) breast composition categories (A-D), (ii) dense (categories C, D) vs. non-dense (categories A, B), and (iii) extremely dense (category D) vs. moderately dense (categories A-C). The models were tested against the reference standard on the independent test set. AI-assisted interpretation was performed by majority voting between the models and each radiologist's assessment. STATISTICAL TESTS: Inter-observer variability was assessed using linear-weighted kappa (κ) statistics. Kappa statistics, accuracy, and area under the receiver operating characteristic curve (AUC) were used to assess models against reference standard. RESULTS: In the independent test set, five readers showed an overall substantial agreement on tasks (i) and (ii), but moderate agreement for task (iii). The best-performing model showed substantial agreement with reference standard for tasks (i) and (ii), but moderate agreement for task (iii). With the assistance of the AI models, almost perfect inter-observer variability was obtained for tasks (i) (mean κ = 0.86), (ii) (mean κ = 0.94), and (iii) (mean κ = 0.94). DATA CONCLUSION: Deep learning and radiomics models have the potential to help reduce inter-observer variability of breast density assessment. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY: Stage 1.

Torsade de pointes arrhythmias arise at the site of maximal heterogeneity of repolarization in the chronic complete atrioventricular block dog.

AIMS: The chronic complete atrioventricular block (CAVB) dog is highly sensitive for drug-induced torsade de pointes (TdP) arrhythmias. Focal mechanisms have been suggested as trigger for TdP onset; however, its exact mechanism remains unclear. In this study, detailed mapping of the ventricles was performed to assess intraventricular heterogeneity of repolarization in relation to the initiation of TdP. METHODS AND RESULTS: In 8 CAVB animals, 56 needles, each containing 4 electrodes, were inserted in the ventricles. During right ventricular apex pacing (cycle length: 1000-1500 ms), local unipolar electrograms were recorded before and after administration of dofetilide to determine activation and repolarization times (RTs). Maximal RT differences were calculated in the left ventricle (LV) within adjacent electrodes in different orientations (transmural, vertical, and horizontal) and within a square of four needles (cubic dispersion). Dofetilide induced TdP in five out of eight animals. Right ventricle-LV was similar between inducible and non-inducible dogs at baseline (327 ± 30 vs. 345 ± 17 ms) and after dofetilide administration (525 ± 95 vs. 508 ± 15 ms). All measurements of intraventricular dispersion were not different at baseline, but this changed for horizontal (206 ± 20 vs. 142 ± 34 ms) and cubic dispersion (272 ± 29 vs. 176 ± 48 ms) after dofetilide: significantly higher values in inducible animals. Single ectopic beats and the first TdP beat arose consistently from a subendocardially located electrode terminal with the shortest RT in the region with largest RT differences. CONCLUSION: Chronic complete atrioventricular block dogs susceptible for TdP demonstrate higher RT differences. Torsade de pointes arises from a region with maximal heterogeneity of repolarization suggesting that a minimal gradient is required in order to initiate TdP.

Verapamil as an antiarrhythmic agent in congestive heart failure: hopping from rabbit to human?

Repolarization-dependent cardiac arrhythmias only arise in hearts facing multiple 'challenges' affecting its so-called repolarization reserve. Congestive heart failure (CHF) is one such challenge frequently observed in humans and is accompanied by altered calcium handling within the contractile heart cell. This raises the question as to whether or not the well-known calcium channel antagonist verapamil acts as an antiarrhythmic drug in this setting, as seen in arrhythmia models without CHF. According to the study of Milberg et al. in this issue of BJP, the answer is yes. The results of this study, using a rabbit CHF model, raise important questions. First, given that the model combines CHF with a number of other interventions that predispose towards arrhythmia, will similar conclusions be reached in a setting where CHF is a more prominent proarrhythmic challenge; second, what is the extent to which other effects of calcium channel block would limit the clinical viability of this pharmacological approach in CHF? In vivo studies in large animal CHF models are now required to further explore this interesting, but complex, approach to the treatment of arrhythmia. LINKED ARTICLE This article is a commentary on Milberg et al., pp. 557-568 of this issue. To view this paper visit http://dx.doi.org/10.1111/j.1476-5381.2011.01721.x.

Effects of K201 on repolarization and arrhythmogenesis in anesthetized chronic atrioventricular block dogs susceptible to dofetilide-induced torsade de pointes.

The novel antiarrhythmic drug K201 (4-[3-{1-(4-benzyl)piperidinyl}propionyl]-7-methoxy-2,3,4,5-tetrahydro-1,4-benzothiazepine monohydrochloride) is currently in development for treatment of atrial fibrillation. K201 not only controls intracellular calcium release by the ryanodine receptors, but also possesses a ventricular action that might predispose to torsade de pointes arrhythmias. The anti- and proarrhythmic effects of K201 were investigated in the anesthetized canine chronic atrioventricular block model. Two doses of K201 (0.1 and 0.3mg/kg/2 min followed by 0.01 and 0.03 mg/kg/30 min i.v.) were tested in 4 serial experiments in dogs with normally conducted sinus rhythm (n=10) and in torsade de pointes-susceptible dogs with chronic atrioventricular block. Susceptibility was assessed with dofetilide (0.025 mg/kg/5 min i.v.). Beat-to-beat variability of repolarization was quantified as short-term variability of left ventricular monophasic action potential duration. In dogs with normally conducted sinus rhythm, both doses of K201 prolonged ventricular repolarization whereas only the higher dose prolonged atrial repolarization. At chronic atrioventricular block, dofetilide induced torsade de pointes in 9 of 10 dogs. K201 did neither suppress nor prevent dofetilide-induced torsade de pointes. K201 dose-dependently prolonged ventricular repolarization. In contrary to the lower dose, the higher dose did increase beat-to-beat variability of repolarization (from 1.2 ± 0.3 to 2.9 ± 0.8 ms, P<0.05) and resulted in spontaneous, repetitive torsade de pointes arrhythmias in 1 of 7 dogs; Programmed electrical stimulation resulted in torsade de pointes in 2 more dogs. In conclusion, both doses of K201 showed a class III effect. No relevant antiarrhythmic effects against dofetilide-induced torsade de pointes were seen. Only at the higher dose a proarrhythmic signal was observed.