AI-Supported Autonomous Uterus Reconstructions: First Application in MRI Using 3D SPACE with Iterative Denoising

• DOI: 10.1016/j.acra.2023.09.035
• Publication Date: 2023-11-03T09:16:17Z
• Authors: Daniel Hausmann; Aline Lerch; Sebastian Hitziger; Monika Farkas; Elisabeth Weiland; Andreas Lemke; Maximilian Grimm; Rahel A. Kubik-Huch
• Keywords: 3D ultrasound

Rationale and ObjectivesT2-weighted imaging in at least two orthogonal planes is recommended for assessment of the uterus. To determine whether a convolutional neural network-based algorithm could be used re-constructions uterus axes derived from 3D SPACE with iterative denoising.Materials Methods50 patients aged 18–81 (mean: 42) years who underwent an MRI examination participated voluntarily this prospective study after informed consent. In addition to standard pelvis protocol, research application sequence was acquired sagittal orientation. Reconstructions both cervix cavum short long were performed by trainee (T), experienced radiologist (E), prototype software (P). next step, reconstructions evaluated anonymously readers according 5-point-Likert-Scales. addition, length cervical canal, distance between tube angles measured on all reconstructions. Interobserver agreement assessed ratings.ResultsFor axes, significant differences found scores T, E P. P received higher preferred significantly more often exception comparison reconstruction Cervix (Cervix short: vs. T: p = 0.02; E: 0.26; long: 0.01; < Cavum 0.01). Regarding diameters, (length canal/cavum/distance angles) larger diameters recorded compared T (mm): 25.43; 25.65; P: 26.65; 26.24; 25.04; 27.33; 31.98; 32.91; 34.41; 0.01); 0.04). Moderate substantial Reader 1 2 (range: 0.39–0.67).ConclusionP able reconstruct as well or better than T. thereby lead workflow facilitation enable efficient reporting uterine MRI. T2-weighted denoising. 50 ratings. For 0.39–0.67).