Automatic Retinoblastoma Screening and Surveillance Using Deep Learning

Abstract

Retinoblastoma is the most common intraocular malignancy in childhood. With the advanced management strategy, the global salvage and overall survival have significantly improved, which proposes subsequent challenges regarding long-term surveillance and offspring screening. Here, we developed deep learning algorithm, called Deep Learning Assistant for Retinoblastoma (DLA-RB) training on A total of 36623 images from 713 patients. We validated it in the prospectively collected dataset, comprised of 1366 images form 139 eyes of 103 patients. In identifying active retinoblastoma tumors (treatment required) from all clinical-suspected patients, the area under the receiver operating characteristic curve (AUC) of DLA-RB reached 0.991 (95% CI 0.970-1.000). In identifying active retinoblastoma from stable retinoblastoma patients (treatment is not required), AUC of DLA-RB reached 0.962 (95% CI 0.915-1.000), respectively. Cost-utility analysis revealed that DLA-RB based diagnosis mode is more cost-effective in both retinoblastoma diagnosis and retinoblastoma activity surveillance. The DLA-RB achieved high accuracy and sensitivity in identifying active retinoblastoma from the normal and stable retinoblastoma fundus. It can be incorporated within telemedicine programs in the future.

Explainable AI for Retinoblastoma Diagnosis: Interpreting Deep Learning Models with LIME and SHAP

“…The comparative analysis of various studies in Table 2 reveals that our study is highly competitive in terms of performance metrics such as accuracy, precision, recall, and F1 score. None of the studies in the table, except Zhang et al [60], utilized any XAI techniques. On the other hand, our study used two popular XAI techniques, LIME and SHAP, for the interpretation of the classification model.…”

“…However, all of these studies mainly focus on detection and segmentation. Zhang et al [60] developed the Deep Learning Assistant for Retinoblastoma (DLA-RB), which uses uses explainable AI to generate visualizations by Grad-CAM to highlight the regions of an image that are most important in the algorithm's prediction. Statistical analysis was conducted using both R-Statistical Software (version 4.1.1, R Foundation for Statistical Computing, Vienna, Austria) and Stata (version 17.0, StataCorp LLC, College Station, TX, USA).…”

Key component performance analysis of lifting spreader automatic unhooking based on Ansys Workbench

“…But most of the large lifting operations require manual assistance or rely on the crane to complete [2] , while the volume and quality of the larger lifting objects requiring only manual assistance cannot complete the work of unhooking. Domestic automatic unhooking is currently primarily automatic air synchronous unhooking to enable air-throwing.Wei Zhang, Mingyi Li, Mingpo Zhou and other people designed the air automatic synchronous unhooking crane design [3] . However, It does not explain how it ensures that multiple ropes can fall off at the same time, which cannot be impossible for large lifting objects.…”