⚙
AI-generated deep summary
by qwen3.7-flash, 2026-08-12
· read from full text
ⓘ
This study developed a deep learning framework to detect endometriotic lesions in laparoscopic images, addressing the challenges of early diagnosis due to non-specific symptoms. The model utilized Attention Based DenseNet for feature extraction and employed the Whale Shark Algorithm for efficient feature selection, with classification performed by a Parallel Depthwise Separable CNN optimized via Northern Goshawk Optimization. Preprocessing involved block filtering to reduce noise while preserving edges, resulting in high performance metrics including 99.10% accuracy on the Roboflow Universe Gynecological Endometriosis Dataset. This paper is centrally about endometriosis — specifically the use of advanced deep learning models for automated detection of endometriotic lesions in medical imaging.
Abstract
Endometriosis is hard to diagnose at an early stage because of non-specific symptoms and the lack of useful non-invasive tests, and often delays treatment and compromises the quality of life. No automated system currently reliably deals with the recognition of endometriotic lesions at an early stage during laparoscopic images. This study was developed to create a deep learning framework based on the Roboflow Universe Gynecological Endometriosis Dataset. Block filtering was used for preprocessing to reduce the noise while preserving edges. Feature extraction Attention Based DenseNet was used for highlighting salient areas of lesions. The Whale Shark Algorithm was used to optimize the feature selection and the classification was done with a Parallel Depthwise Separable CNN and hyperparameters were optimized using Northern Goshawk Optimization. The Whale Shark Algorithm does efficient feature selection, i.e. keeping biologically relevant features (reducing dimensionality). Classification uses a Parallel Depthwise Separable Convolution Neural Network with hyperparameters optimized using Northern Goshawk Optimization for better performance in terms of accuracy and efficiency. On the Roboflow universe dataset, the proposed model achieves an accuracy of 99.10%, and 98.80% precision, 98.65% recall, and 98.72%, an F1-score. These results provide evidence of a promising, practical tool that can help clinicians to predispose patients with early-stage endometriosis and potentially early intervention for better patient outcomes.
Full text
1,843 characters
· extracted from
oa-doi-fallback
· click to expand
A Novel Endometriosis Detection Using Advanced Deep Learning Model with Whale Shark Algorithm
DOI:
https://doi.org/10.71086/IAJSE/V12I4/IAJSE1294Keywords:
Endometriosis Detection, Whale Shark Algorithm (WSA), Roboflow Universe Dataset, Feature Selection Optimization, Gynecology, ATT Densenet.Abstract
Endometriosis is hard to diagnose at an early stage because of non-specific symptoms and the lack of useful non-invasive tests, and often delays treatment and compromises the quality of life. No automated system currently reliably deals with the recognition of endometriotic lesions at an early stage during laparoscopic images. This study was developed to create a deep learning framework based on the Roboflow Universe Gynecological Endometriosis Dataset. Block filtering was used for preprocessing to reduce the noise while preserving edges. Feature extraction Attention Based DenseNet was used for highlighting salient areas of lesions. The Whale Shark Algorithm was used to optimize the feature selection and the classification was done with a Parallel Depthwise Separable CNN and hyperparameters were optimized using Northern Goshawk Optimization. The Whale Shark Algorithm does efficient feature selection, i.e. keeping biologically relevant features (reducing dimensionality). Classification uses a Parallel Depthwise Separable Convolution Neural Network with hyperparameters optimized using Northern Goshawk Optimization for better performance in terms of accuracy and efficiency. On the Roboflow universe dataset, the proposed model achieves an accuracy of 99.10%, and 98.80% precision, 98.65% recall, and 98.72%, an F1-score. These results provide evidence of a promising, practical tool that can help clinicians to predispose patients with early-stage endometriosis and potentially early intervention for better patient outcomes.
Text is read by the "Ask this paper" AI Q&A widget below.
Extraction quality varies by source — PMC NXML preserves structure
cleanly, OA-HTML may include some navigation residue, and OA-PDF can
have broken hyphenation. The publisher copy
(via DOI)
is the canonical version.