Advancing retinoblastoma detection based on binary arithmetic optimization and integrated features

Nuha Alruwais; Marwa Obayya; Fuad Al-Mutiri; Mohammed Assiri; Amani A Alneil; Abdullah Mohamed

doi:10.7717/peerj-cs.1681

Advancing retinoblastoma detection based on binary arithmetic optimization and integrated features

PeerJ Comput Sci. 2023 Nov 22:9:e1681. doi: 10.7717/peerj-cs.1681. eCollection 2023.

Authors

Nuha Alruwais¹, Marwa Obayya², Fuad Al-Mutiri³, Mohammed Assiri⁴, Amani A Alneil⁴, Abdullah Mohamed⁵

Affiliations

¹ Department of Computer Science and Engineering, College of Applied Studies and Community Services, King Saud University, Saudi Arabia, Riyadh, Saudi Arabia.
² Department of Biomedical Engineering, College of Engineering, Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia.
³ Department of Mathematics, King Khalid University, Muhayil Asir, Saudi Arabia.
⁴ Department of Computer Science, Prince Sattam Bin Abdulaziz University, Aflaj, Saudi Arabia.
⁵ Research Centre, Future University, New Cairo, Egypt.

Abstract

Retinoblastoma, the most prevalent pediatric intraocular malignancy, can cause vision loss in children and adults worldwide. Adults may develop uveal melanoma. It is a hazardous tumor that can expand swiftly and destroy the eye and surrounding tissue. Thus, early retinoblastoma screening in children is essential. This work isolated retinal tumor cells, which is its main contribution. Tumors were also staged and subtyped. The methods let ophthalmologists discover and forecast retinoblastoma malignancy early. The approach may prevent blindness in infants and adults. Experts in ophthalmology now have more tools because of their disposal and the revolution in deep learning techniques. There are three stages to the suggested approach, and they are pre-processing, segmenting, and classification. The tumor is isolated and labeled on the base picture using various image processing techniques in this approach. Median filtering is initially used to smooth the pictures. The suggested method's unique selling point is the incorporation of fused features, which result from combining those produced using deep learning models (DL) such as EfficientNet and CNN with those obtained by more conventional handmade feature extraction methods. Feature selection (FS) is carried out to enhance the performance of the suggested system further. Here, we present BAOA-S and BAOA-V, two binary variations of the newly introduced Arithmetic Optimization Algorithm (AOA), to perform feature selection. The malignancy and the tumor cells are categorized once they have been segmented. The suggested optimization method enhances the algorithm's parameters, making it well-suited to multimodal pictures taken with varying illness configurations. The proposed system raises the methods' accuracy, sensitivity, and specificity to 100, 99, and 99 percent, respectively. The proposed method is the most effective option and a viable alternative to existing solutions in the market.

Keywords: Deep learning; Ophthalmology; Retinal tumor; Retinoblastoma; Segmentation.

Grants and funding

This study was funded by the Deanship of Scientific Research at King Khalid University for funding this work through large group Research Project under grant number (RGP2/35/44), Princess Nourah bint Abdulrahman University Researchers Supporting Project number (PNURSP2023R203), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia. Research Supporting Project number (RSPD2023R608), King Saud University, Riyadh, Saudi Arabia, Prince Sattam bin Abdulaziz University project number (PSAU/2023/R/1444), and by the Future University in Egypt (FUE). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.