Page 46 - AIH-1-4

Artificial Intelligence in Health Segmentation and classification of DR using CNN

statement emphasizes the challenges in existing diagnostic Writing-original draft: Manoj Saligrama Harisha, Arya
methods and introduces datasets encompassing lesion Arun Bhosale
segmentation, disease/image grading, and LLMs for test/ Writing-review & editing: Manoj Saligrama Harisha, Arya
treatment recommendations. Arun Bhosale
Identifying research gaps, the paper underscores the Ethics approval and consent to participate
need for exploring the integration of pre-trained LLMs
with segmented image data, emphasizing the potential Not applicable.
synergies between visual segmentation features and
clinical classifications within a decision-support system. Consent for publication
Another research gap pertains to the dynamics of multi- Not applicable.
model integration, particularly in a web application context
where lesion segmentations, disease classification, and Availability of data
LLMs collaborate. The objectives outline a comprehensive Data for this study are sourced from two primary datasets:
DR detection methodology, the integration of various the APTOS 2019 Blindness Detection dataset, available
models, performance evaluation, contribution to detection on Kaggle.com, and the IDRiD, accessible through Grad-
methodologies, and the identification and exploration of Challenge.org. In addition, the code repository for the DR
research gaps. Detection and Classification System used in this research

The research scope aims to revolutionize DR detection can be found at https://github.com/Manoj-Sh-AI/Diabetic-
methodologies by integrating cutting-edge technologies Retinopathy-Detection-and-Clasification-System.
and contributing to the wider field of medical imaging
and automated diagnostics. The methods encompass data Further disclosure
pre-processing, data augmentation, network architecture This paper has been uploaded and made publicly available
detailing the U-Net model, and the training and testing on a preprint server (arXiv:2401.02759), available at:
processes. The results showcase the effectiveness of the https://arxiv.org/abs/2401.02759
segmentation model across various retinal structures,
with high Jaccard scores, F1 scores, recall scores, precision References
scores, and overall accuracy, underscoring its potential 1. NCHS. Eye Disorders and Vision Loss among U.S. Adults
for enhancing diagnostic capabilities in retinal pathology Aged 45 and Over with Diagnosed Diabetes; 2019. Available
assessment. from: https://stacks.cdc.gov/view/cdc/80081 [Last accessed
on 2024 Jul 11].
Acknowledgments
2. Pezzullo L, Streatfeild J, Simkiss P, Shickle D. The economic
We would like to express our gratitude to our guide, impact of sight loss and blindness in the UK adult population.
Dr. Narender M., Assistant Professor, Department of BMC Health Serv Res. 2018;18:63.
Computer Science and Engineering at the National doi: 10.1186/s12913-018-2836-0.
Institute of Engineering (NIE), for his invaluable guidance
and support throughout the course of this research. His 3. Hann CE, Chase JG, Revie JA, Hewett D, Shaw GM. Diabetic
expertise and insightful feedback significantly contributed retinopathy screening using computer vision. IFAC Proc
to the development of this work. Vol. 2009;42:298-303.
doi: 10.3182/20090812-3-DK-2006.0086
Funding
4. Rohan TE, Frost CD, Wald NJ. Prevention of blindness
None. by screening for diabetic retinopathy: A quantitative
assessment. BMJ. 1989;299:1198-1201.
Conflict of interest
doi: 10.1136/bmj.299.6709.1198
The authors declare that they have no competing interests. 5. SNEC. Singapore’s Eye Health; 2019. Available from: https://
www.snec.com.sg [Last accessed on 2024 Jul 11].
Author contributions
6. Priya R, Aruna P. SVM and Neural Network Based Diagnosis
Conceptualization: Manoj Saligrama Harisha, Arya Arun of Diabetic Retinopathy; 2012. Available from: https://www.
Bhosale researchgate.net/publication/261177114_svm_and_neural_
Investigation: All authors network_based_diagnosis_of_diabetic_retinopathy [Last
Methodology: Manoj Saligrama Harisha accessed on 2024 Jul 11].

Volume 1 Issue 4 (2024) 40 doi:10.36922/aih.2783