Developing a deep learning model to predict the breast implant texture types with ultrasonography image: feasibility study

Introduction Breast implants, including textured variants, have been widely used in aesthetic and reconstructive mammoplasty. However, the textured type, which is one of the shell types of breast implants, has been identified as a possible carcinogenic factor for lymphoma, specifically breast implant-associated anaplastic large cell lymphoma (BIA-ALCL). Identifying the texture type of the implant is critical to the diagnosis of BIA-ALCL. However, distinguishing the shell type can be difficult due to human memory or loss of medical history. An alternative approach is to use ultrasonography, but this method also has limitations in quantitative assessment. Objective The objective of this study is to determine the feasibility of using a deep learning model to classify the textured shell type of breast implants and make robust predictions from ultrasonography images from heterogeneous sources. Methods A total of 19,502 breast implant images were retrospectively collected from heterogeneous sources, including images from both Canon (D1) and GE (D2), images of ruptured implants (D3), and images without implants (D4), as well as publicly available images (D5). The Canon (D1) images were trained using Resnet-50. The performance of the model on D1 was evaluated using stratified 5-fold cross-validation. Additionally, external validation was conducted using D2 and D5. The AUROC and PRAUC were calculated based on the contribution of the pixels with Grad-CAM. To identify the significant pixels for classification, we masked the pixels that contributed less than 10%, up to a maximum of 100%. To assess model robustness to uncertainty, Shannon entropy was calculated for four image groups: Canon (D1), GE (D2), ruptured implant (D3), and without implants (D5). Result The deep learning model achieved an average AUROC of 0.98 and a PRAUC of 0.88 in the Canon dataset (D1). For images captured with GE (D2), the model achieved an AUROC of 0.985 and a PRAUC of 0.748. Additionally, the model predicted an AUROC of 0.909 and a PRAUC of 0.958 for a dataset available online. For quantitative validation, this model maintained PRAUC up to 90% masking of less contributing pixels, and the remnant pixels located in breast shell layers. Furthermore, the prediction uncertainty increased in the following order: Canon (D1), GE (D2), ruptured implant (D3), no implant (D5) (0.066; 0072; 0.371; 0.777, respectively). Conclusion We have demonstrated the feasibility of using deep learning to predict the shell types of breast implants. With this approach, the textured shell types of breast implants can be quantified, supporting the first step in the diagnosis of BIA-ALCL. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement None of the authors has a financial interest in any products, devices, or drugs mentioned in this manuscript. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: This retrospective study was approved by the Internal Institutional Review Board of the Korea National Institute of Bioethics Policy (IRB No. P01-202401-01-006), which waived the requirement for informed consent of medical records, including patients' images and characteristics I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes This dataset is available in .