SWOT Analysis of Vision Transformers (ViTs) for Automated Diagnosis of Endometriosis from Laparoscopic Videos: Feasibility and Ethical Challenges

Automated analysis of laparoscopic videos is emerging as a valuable tool for improving the detec- tion of diseases such as endometriosis, offering the potential to assist surgeons with complex in- traoperative decisions and to standardize diagnostic accuracy [1, 2]. Traditional deep learning approaches for surgical video analysis have relied heavily on convolutional neural networks (CNNs) and recurrent neural networks (RNNs), which excel at capturing local spatial features and 1. Department of Gynecology, Babol University of Medical Sciences, Babol, Iran; 2*. Corresponding author: Department of Surgery, Shahid Beheshti University Medical Sciences, Tehran, Iran. Email: [email protected] ; 3. Faculty of Medicine, Mazandaran University of Medical science, Mazandaran, Iran.4. Student Research Committee, Semnan Uni- versity of Medical Sciences, Semnan, Iran; 5. Student Research Committee, Babol University of Medical Sciences, Babol, Iran. / Open Access. © 202 5 the author(s), published by InfoPub. This work is licensed under the Creative Commons Attribution 4.0 International License. (Journal homepage: https://www.isjtrend.com) https://doi.org/10.61186/ist.202502.05.02 Endometriosis diagnosis via laparoscopy remains challenging due to subtle lesion appearances and in- ter-observer variability. While artificial intelligence shows promise for surgical video analysis, the po- tential of Vision Transformers (ViTs) specifically for endometriosis detection remains unexplored. This study applied a SWOT framewor k to evaluate ViTs for automated endometriosis diagnosis in laparo- scopic videos. Analysis of 10 studies from PubMed, IEEE Xplore, and Scopus identified key findings: Strengths included (1) global attention for lesion detection, (2) outperforming CNNs/RNNs in surgical tasks (91-97% accuracy), and (3) multimodal data integration. Weaknesses were (1) dependence on unavailable annotated datasets, (2) high computational needs, (3) limited local feature sensitivity, and (4) annotation variability issues. Opportun ities involved (1) self -supervised learning from unlabeled videos and (2) explainable attention maps. Threats comprised (1) performance variability across surgi- cal settings, (2) lacking regulatory standards, and (3) data privacy concerns. Crucially, no stu dies di- rectly tested ViTs for endometriosis diagnosis despite their potential. For clinical implementation, three requirements emerged: (1) collaborative dataset creation, (2) optimized hybrid architectures, and (3) ethical guidelines for surgical AI. This structured analysis provides a roadmap for developing ViT-based diagnostic tools while addressing current limitations in data, technology, and clinical integration . Vision Transformers (ViTs), Endometriosis, Laparoscopic Surgery, SWOT Analysis, Ethical AI. InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com short-term temporal patterns but often struggle with the nuanced and diffuse presentations typ- ical of gynecologic pathologies like endometriosis [3, 4]. Advances in Vision Transformers (ViTs) and transformer -inspired architectures have introduced a new paradigm capable of modeling global spatial and temporal relationships in video data, resulting in superior performance over conventional models for tool detection, workflow analysis, and related tasks within laparoscopy [2, 5]. Despite this progress, the application of ViT -based models to direct disease diagnosis from laparoscopic video, and specifically to endometriosis identification, remains largely uncharted in the published literature [6, 7]. The feasibility of deploying such models is challenged by factors including data scarcity, high annotation costs, and significant computational demands required for real-time clinical integration [4, 8]. Moreover, ethical issues such as annotation burden, data privacy, interpretability, and regulatory acceptance have not been systematically explored in the context of transformer-based surgical video analysis [7, 9]. To systematically assess these complex dimensions, the SWOT (Strengths, Weaknesses, Op- portunities, Threats) analysis framework is increasingly used in technology evaluation [10]. SWOT provides a structured approach to examining both the internal capabilities and limitations of a technology, as well as the external factors that may facilitate or hinder its clinical adoption [11]. By identifying strengths and weaknesses intrinsic to ViT architectures, and analyzing oppor- tunities and threats rooted in clinical, ethical, or regulatory contexts, a SWOT analysis can offer a comprehensive perspective for guiding future research and practical implementation. The purpose of this study is to bridge these critical gaps by providing a structured SWOT anal- ysis of Vision Transformer models for the automated diagnosis of endometriosis from laparo- scopic videos. This review aims to evaluate the current state of technical feasibility and to high- light the ethical challenges involved, offering practical considerations for future research and clin- ical translation.

Study Design This study employs a qualitative, structured SWOT (Strengths, Weaknesses, Opportunities, Threats) analysis to systematically evaluate the feasibility and ethical challenges of deploying Vi- sion Transformer (ViT) architectures for automated diagnosis of endometriosis from laparoscopic video data. Literature Search and Data Sources A comprehensive literature search was conducted using PubMed, IEEE Xplore and Scopus to identify primary research articles, systematic reviews, and technical reports relevant to: • Vision Transformer (ViT) or transformer-inspired models in medical image/video analy- sis • Automated disease diagnosis in laparoscopy, including but not limited to endometriosis • Feasibility, clinical integration, and ethical aspects of AI in surgical video analysis Search terms included combinations of Technology Terms: " vision transformer" OR "ViT" OR "visual transformer" OR "transformer architecture" OR "attention mechanism". Clinical Application Terms: "laparoscopic video" OR "surgical video" OR "endoscopic video" OR "minimally invasive sur- gery". Diagnosis Terms: "disease diagnosis" OR "pathology detection" OR "lesion identification" OR InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com "medical diagnosis" .Analysis Framework Terms: "technology assessment" OR "clinical implemen- tation" OR "adoption barriers". Ethical Considerations: "ethical challenges" OR "AI ethics" OR "al- gorithmic bias", and related synonyms. Inclusion and Exclusion Criteria Inclusion criteria • Peer-reviewed articles or preprints published in English • Studies describing ViT or transformer -based machine learning approaches in laparo- scopic video analysis • Papers addressing automated disease diagnosis, particularly for endometriosis, or dis- cussing general feasibility and ethical considerations for AI in surgical video Exclusion criteria • Studies focusing solely on surgical phase recognition, tool detection, or workflow analysis without disease/pathology diagnosis • Articles without explicit discussion of model strengths, weaknesses, opportunities, or threats SWOT Framework Development Drawing on the included literature and expert domain knowledge, a SWOT matrix was con- structed to capture: • Strengths: Inherent capabilities and advantages of ViT models for disease detection in laparoscopic video • Weaknesses: Technical and practical limitations, including data and clinical constraints • Opportunities: External factors and future directions that may facilitate clinical adop- tion, innovation, or improved outcomes • Threats: Risks, barriers, and ethical concerns associated with real -world deployment and broader societal implications Evaluation Process Each article was independently reviewed by two researchers. Factual statements regarding ViT architectures, feasibility, and ethical issues were extracted and categorized into the SWOT components. Discrepancies were resolved through discussion.

Systematic Literature Search and Screening Process Our comprehensive search across PubMed (n=45), IEEE Xplore (n=32), and Scopus (n=28) initially identified 105 articles published between 2018 –2023. After removing 23 duplicates, 82 unique records underwent title/abstract screening. Of these, 52 were excluded for irrelevance to laparoscopic video analysis or lack of focus on Vision Transformers (ViTs). The remaining 30 full-text articles were assessed against predefined inclusion/exclusion cri- teria. 20 additional studies were excluded for the following reasons: • 12 studies focused exclusively on surgical tool detection or workflow segmentation without disease diagnosis. • 7 studies employed CNNs/RNNs only, omitting ViT or transformer-based architectures. • 1 study was excluded due to the non-English language. InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com The final 10 articles (Table 1) met all criteria, including: • Use of ViT or hybrid transformer models in laparoscopic/endoscopic video analysis. • Relevance to automated disease diagnosis (though none addressed endometriosis directly). • Discussion of feasibility, clinical integration, or ethical challenges. Notably, PubMed contributed 6 studies, IEEE Xplore 3, and Scopus 1, reflecting the interdis- ciplinary nature of the field (computer science vs. clinical journals). Table 1. Summary of Included Studies on Vision Transformers (ViTs) in Surgical Video Analysis. Study Purpose Type Method &

Results Conclusion [9] Develop and assess a multimodal transformer model for analyzing laparoscopic surgical videos, aiming to reduce negative outcomes and improve patient safety. Original research (empirical) Multimodal model inspired by Video- Audio-Text Transformer; uses ViT for images and BERT for text; trained on Cholec80 LC videos with various complexities. Mean accuracy 91.0%, precision 81%, recall 83% on 30/80 test videos (Cholec80 dataset). Shows the model can extract hidden and distinct features, helping to create safer surgery systems, but the overall role and advantages of AI models in surgery remain uncertain. [12] Propose dataset and

for event recognition in laparoscopic gynecology videos; evaluate hybrid transformer for detecting critical intra- and post- operative events. Original research (empirical) Introduces annotated event dataset; compares several CNN-RNN models; develops hybrid transformer architecture for event recognition; uses frame sampling. The hybrid transformer improves recognition accuracy, counteracts occlusion/motion blur, and yields high temporal resolution in event recognition. The proposed hybrid transformer approach is superior to existing CNN-RNN methods for event recognition in laparoscopic surgery videos. [2] Detect surgical tool presence in laparoscopic video using a transformer architecture (LapFormer). Original research (empirical) LapFormer model: feed-forward transformer with attention for inter- frame correlation; evaluated on Cholec80 dataset. Outperforms CNN and RNN baselines by 20.3 and 17.3 points, respectively, in macro-F1 score; includes ablation studies. Transformer architecture is more effective than previous methods for surgical tool detection in laparoscopic videos. [7] Explore use of pure vision transformers for classifying single- and multi-label surgical tool frames in laparoscopic surgery. Original research (empirical) Pure ViT models for SL/ML tool classification; 5- fold cross- validation on Cholec80 dataset. Mean average precision (mAP) = 95.8%, outperforming conventional multi- label models.

suggest promise for ViT models in surgical tool detection, warranting further research. [13] Present a bidirectional transformer with sparse attention for GI disease recognition from endoscopy images. Original research (empirical) Bidirectional Transformer with Sparse Attention (BTSA); trained and tested on large-scale GI endoscopy datasets. BTSA achieves outstanding performance, surpasses existing models in GI disease recognition, and is efficient. Model shows significant potential, but further research and validation are needed for clinical utility. [8] Benchmark deep learning models (convolutional, transformer, hybrid) for surgical workflow/phase Original research (empirical) Compares fully convolutional, fully transformer, and hybrid models; workflow recognition from Hybrid model achieves 93% frame- level accuracy and 85 segmental edit distance; fully transformer also gets Hybrid models effectively capture workflow and yield best results in surgical video analysis. InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com recognition in endoscopic videos. gastric bypass surgery videos. state-of-the-art results. [14] Develop a CNN with a new attention module (P-CSEM) to improve laparoscopic tool detection. Original research (empirical) CNN with P-CSEM attention modules at various layers; trained and tested on Cholec80 dataset. Attention-enhanced model achieves mean average precision of 93.14%; feature relevance improved. Custom attention modules inside CNNs can enhance tool relevance in laparoscopic tool classification. [15] Use a transformer- based, multi-view approach to improve surgical phase recognition in laparoscopic cholecystectomy. Original research (empirical) Multi-view phase recognition using transformer-based model with late fusion of laparoscopic and in-room camera data. Performance is mixed; real-world data collection introduced challenges; model performance decreased with poor data. Integration of multi-view data is complex; real- world diversity and better data are needed for optimal models. [16] Review and summarize deep learning methods (incl. transformers) for phase and step recognition in surgical workflow analysis. Systematic review (review article) Systematic review; searched databases for studies post-2018 on deep learning

for surgical workflow; 44 studies reviewed. Temporal context (RNN, CNN, Transformers) is key; lack of diverse datasets is a major challenge in workflow recognition. The field is advancing, but robust, generalizable models are hampered by limited, supervised datasets. [4] Compare performance of ViT, CNN, and hybrid CNN-ViT architectures for GI disease classification from endoscopy images. Original research (empirical) ViT, hybrid CNN- ViT, CNN architectures; trained on GI endoscopy images (WCE, etc.); classified 6 GI disease classes. Hybrid model achieves test accuracy of 97.91%, F1 97.91%, precision 98.01%; compared against CNN and ViT models. Hybrid CNN-ViT models can accurately classify GI diseases from endoscopy images. Vision Transformer (ViT) or Transformer-Inspired Models in Medical Image/Video Analysis Recent years have seen increased application of Vision Transformer (ViT) and transformer - inspired architectures in medical image and video analysis. Multiple studies have focused on lev- eraging the global attention mechanisms of ViTs and their hybrid forms for diverse tasks, includ- ing tool detection, workflow and phase recognition, and disease diagnosis in gastrointestinal en- doscopy images. These models have shown performance improvements over traditional CNNs or RNNs, especially for spatio-temporal data in laparoscopic and endoscopic domains. However, the predominant use cases reported in the literature relate to surgical tool or workflow recognition, with relatively few studies targeting disease diagnosis, and none specifically focusing on endome- triosis in laparoscopic videos [4, 7-9, 12] (Table 2). Table 2. Applications of ViTs in Laparoscopic/Endoscopic Video Tasks. Study Task Data Type ViT Usage Key Outcome [3] Surgical tool detection Laparoscopic video Pure ViT Superior macro-F1 to CNN/RNN (+20%) [7] Tool classification (SL/ML) Laparoscopic video Pure ViT mAP 95.8%; outperforming previous CNN models [8] Workflow/phase recognition Laparoscopic video ViT/CNN/Hybrid Transformers match SOTA convolutional results [4] GI disease classification Capsule endoscopy img ViT/Hybrid CNN ViT and hybrids outperform CNN only [13] GI disease recognition GI endoscopy images Bi-directional T Sparse attention improves efficiency InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com Automated Disease Diagnosis in Laparoscopy, Including but Not Limited to Endometriosis Automated disease diagnosis in laparoscopy remains a relatively underexplored application compared to tool detection or workflow recognition. Across the reviewed literature, there are no studies directly addressing endometriosis detection with ViT -based methods from laparoscopic videos. While several references evaluate transformers for disease classification in gastrointesti- nal endoscopy images, these studies do not utilize laparoscopic video data nor focus on gyneco- logical pathologies such as endometriosi s. The only disease diagnosis applications identified re- late to GI pathologies from endoscopy, not laparoscopy [4]. Tool and event recognition studies in laparoscopy do not provide systematic assessment of disease classification performance [7-9, 12] (Table 3). Table 3. Gap Analysis: Disease Diagnosis in Laparoscopy vs. Endoscopy. Study Disease Task Method Laparoscopic Video Outcome/Limitations [13] GI disease (general) Bi-directional T No Applies to endoscopy, not laparoscopy [4] GI disease (6 types) ViT/Hybrid CNN No Only WCE images, not laparoscopy video [2, 7-9, 12] Tool/phase/event ViT/Hybrid/CNN Yes No disease/pathology classification task Feasibility, Clinical Integration, and Ethical Aspects of AI in Surgical Video Analysis Feasibility and clinical integration of AI models—especially Vision Transformers—in surgical video analysis are discussed peripherally in several empirical studies, though not as primary out- comes. The main technical feasibility concerns center on data scar city, need for large and granu- larly annotated training datasets, and high computational demands for real-time application. For clinical integration, general requirements include high interpretability (attention -based heatmaps), robust performance across ha rdware/sites, and alignment with surgical workflow. Ethical aspects, such as annotation burden, trust in AI, and concerns over robust deployment, are noted in generic terms in the literature, with no detailed frameworks specific to ViT-based disease diagnosis in laparoscopy [4, 7-9]. Safety, explainability, and regulatory barriers are highlighted as ongoing challenges (Table 4). Table 4. Key Feasibility and Ethical Challenges for ViT Deployment. Aspect Noted Issues Studies Citing the Issue Data hunger Requires large annotated datasets [2, 4, 7-9] Computation Resource-intensive; real-time limits [2, 7-9] Interpretation Needs explainable attention maps [2, 9] Annotation cost High labeling burden [2, 7, 8] Ethical/safety Clinical trust, regulatory hurdles [2, 4, 7, 8] SWOT Analysis Strengths of ViT for Automated Diagnosis of Endometriosis from Laparoscopic Videos Vision Transformers provide a global receptive field allowing for comprehensive aggregation of visual information across each frame and, in extended forms, across video sequences. This spa- tial (and potentially temporal) attention facilitates detection of diffuse, subtle, or irregularly shaped disease lesions commonly seen in endometriosis. ViTs are also highly adaptable to multi- modal fusion, supporting the integration of additional context such as surgical reports or instru- ment data. Empirical studies in tool and workflow recognition indicate ViTs can outperform tra- ditional CNNs, particularly when applied to well-annotated surgical video datasets [2, 7-9]. InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com Weaknesses of ViT for Automated Diagnosis of Endometriosis from Laparoscopic Videos A primary weakness is the requirement for large quantities of diverse, accurately annotated video data for effective model training—resources scarce in laparoscopic disease diagnosis, espe- cially for conditions like endometriosis. ViTs are computationally d emanding, posing challenges for real-time operating room integration without significant hardware support or model optimi- zation. Their limited inherent inductive bias for local features can reduce sensitivity to small or localized lesions, unless modified with hybrid architectures. Issues with label quality and interob- server variability in disease annotation may also affect reliability [2, 7-9]. Opportunities of ViT for Automated Diagnosis of Endometriosis from Laparoscopic Videos Emerging opportunities include leveraging self-supervised pretraining on large, unannotated laparoscopic video archives to reduce manual labeling requirements and improve model robust- ness. Federated learning across multiple centers can help build generaliz able models while pre- serving patient privacy. Explainable AI techniques leveraging visual attention maps may facilitate clinical acceptance by enhancing model interpretability. Additionally, advances in spatio -tem- poral transformer variants and multimodal a pproaches could further enhance disease detection capabilities [2, 4, 7-9]. Threats of ViT for Automated Diagnosis of Endometriosis from Laparoscopic Videos Key threats include vulnerability to domain shifts caused by different surgical equipment, camera systems, lighting conditions, and operator techniques —all of which could degrade ViT model performance if not addressed. The lack of standardized datasets and high-quality labeled examples for endometriosis remains a significant barrier. Further, regulatory approval for clinical deployment may be hindered by the complexity and opacity of transformer models, challenges with real-time guarantees, and persistent c oncerns about trust and accountability in automated decision-making [2, 4, 7, 8]. SWOT Matrix for ViT-Based Endometriosis Diagnosis from Laparoscopic Videos The SWOT matrix synthesizes internal (Strengths/Weaknesses) and external (Opportuni- ties/Threats) factors influencing ViT -based endometriosis diagnosis. Strengths highlight ViTs’ global attention for detecting diffuse lesions, while weaknesses address data/computational de- mands. Opportunities include federated learning and explainable AI, whereas threats encompass domain shifts and regulatory barriers. This framework guides balanced evaluation for clinical adoption (Figure 1). InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com Figure 1. SWOT Matrix for ViT-Based Endometriosis Diagnosis from Laparoscopic Videos. Feasibility and Ethical Challenges The use of Vision Transformer (ViT) architectures for automated disease diagnosis in laparo- scopic videos offers significant promise but also faces considerable challenges (Table 5). Moreo- ver, integrating ViT-based diagnostic models into clinical laparoscopic workflows introduces sev- eral ethical concerns, many of which remain unresolved in the literature. A key challenge is the annotation burden: developing large, high -quality labeled datasets for surgical videos demands substantial time from medical experts and raises potential risks regarding patient data privacy [2, 8] (Table 5). Table 5. Feasibility and Ethical Challenges for ViT-based Automated Diagnosis of Endometriosis from Laparoscopic Videos. Domain Challenges/Considerations References Feasibility - Large, annotated laparoscopic video datasets are scarce, especially with disease/pathology labels (e.g., endometriosis). [2, 4, 7, 8] - ViTs require substantial computational resources; real-time clinical OR deployment is challenging. [7, 8] - Annotation requires expert clinicians, leading to high costs and annotation burden. [2, 7, 8] - Self-supervised pretraining and federated learning proposed to address data scarcity and improve generalization but are not yet standard in this context. [4, 9] - Hybrid or sparse attention architectures may partially alleviate computational loads; further validation is needed. [8, 13] - No direct empirical evidence exists for ViT feasibility in endometriosis diagnosis; existing evidence is from adjacent tasks (tool/phase detection). [2, 4, 7, 9] Ethical - High annotation burden and privacy concerns associated with creation and use of surgical video datasets. [2, 7, 8] Challenges - Potential for bias (class imbalance, label noise, interobserver variability) may compromise fairness and reliability. [4, 8, 16] - "Black-box" nature of ViTs limits interpretability; attention maps help but have limits. [2, 9] - Regulatory, accountability, and liability issues are more complex due to ViT model opacity and performance uncertainty. [7, 8] - Consent, secure data storage, and transparency are unresolved in current frameworks for AI in surgical video analysis. [2, 4, 7, 8] InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com Implications The absence of direct evidence for ViT -based endometriosis diagnosis underscores a critical gap in both the research landscape and clinical translation efforts for advanced AI models in gy- necologic laparoscopy. While the strengths of transformers —such as global attention, flexibility for integrating multiple data modalities, and state-of-the-art performance—present clear oppor- tunities for improving disease detection, significant weaknesses and external threats must be ad- dressed. These include the need for large -scale, high-quality annotated datasets, computational barriers to real-time deployment, vulnerability to domain shift, and lack of explainability required for surgeon trust and regulatory approval. The practical integration of ViT-based diagnostic AI in surgical settings therefore depends not only on further technological innovation, but also on ro- bust clinical validation, ethical guidelines, and cross-disciplinary collaboration. Recommendations for Practice 1. Data Infrastructure and Sharing Collaborative efforts should be initiated to develop, standardize, and share large, annotated laparoscopic video datasets—specifically including cases of endometriosis—to enable model training and robust benchmarking [2, 7, 8]. 2. Model Development and Validation Researchers should explore hybrid and spatio- temporal transformer architectures, leveraging approaches such as self-supervised and federated learning to address data scarcity and improve cross-site generalizability [2, 4, 7-9]. 3. Interpretability and Trust Development of explainable AI techniques—such as attention heatmaps—should be prioritized to support clinical interpretation and facilitate trust among surgeons and regulatory bodies [2, 9]. 4. Ethical and Regulatory Oversight Early integration of ethical considerations, including annotation burden, bias, and informed consent, is essential. Engagement with regulatory agencies should guide the development of compliance-ready AI systems [2, 4, 7, 8]. 5. Clinical Integration Pilot implementation of ViT-based diagnostic support tools should be closely monitored in controlled settings, ensuring that workflow integration, computational requirements, and real-time performance meet clinical and patient safety standards [8].

This study provides the first structured SWOT analysis of Vision Transformer (ViT) models for automated diagnosis of endometriosis from laparoscopic videos —a task that remains un- addressed directly in the current literature. While recent advances in deep learning have estab- lished ViT and transformer -inspired architectures as state -of-the-art tools for complex visual tasks, a comparative examination of their performance, limitations, and broader implications in surgical video analysis is essential for informed development and adoption. Compared to prior convolutional and recurrent approaches, ViT models show clear empirical strengths in related laparoscopic video analysis tasks. For instance, in tool detection and work- flow/phase recognition, ViT -based and hybrid transformer models achiev ed superior accuracy and macro-F1 scores over conventional CNN and RNN baselines [2, 7, 8]. These gains stem largely from the global attention mechanisms that allow ViTs to model spatial dependencies and long - range context more efficiently than convolutional filters or sequential RNN layers. Additionally, ViT architectures are highly adaptable: studies show their applicability to single- and multi-label InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com classification [7], as well as their extension to multimodal fusion, incorporating text and audio data for richer surgical scene understanding [9]. However, in direct comparison, the domain of automated disease diagnosis—particularly for conditions like endometriosis—is far less developed for ViT -based models than the adjacent do- mains of surgical tool recognition or workflow segmentation. While transf ormer models have been applied with promising results to GI disease diagnosis from endoscopic images [4], these studies do not use laparoscopic video data, nor do they address gynecological pathologies specif- ically. Tool detection and workflow studies in laparoscopy [2, 7-9, 14, 15] consistently exclude disease classification as an explicit outcome. This gap is striking, as it suggests that the advantages observed for ViT models in instrument and phase tasks may not translate automatically to disease recognition in the more heterogeneous and visually subtle context of endometriosis. Feasibility constraints form a recurrent theme across the literature. Data scarcity and annota- tion costs are reported as the primary bottlenecks for transformer training in laparoscopy [2, 7]. ViT models require large datasets —often orders of magnitude greater than those needed for CNNs—to generalize well and avoid overfitting. This demand is particularly challenging in endo- metriosis research, where annotated laparoscopic video collections are uncommon. Computa- tional barriers also persist, with studies highlighting the significant memory and processing re- quirements of self-attention mechanisms, especially for long surgical videos [4,6]. In comparative terms, these challenges exceed those faced by most CNN/RNN -based pipelines and necessitate innovations such as hybrid architectures, sparse attention mechanisms, or pretraining with large- scale unlabeled data [4, 9, 13]. Ethically, the transformer literature underlines issues that are not unique to ViT models but are amplified by their complexity. Class imbalance, label noise, and interobserver variability — ubiquitous in medical annotation—pose risks for fairness and reliability of automated disease di- agnosis [4, 8]. Where prior CNN/RNN approaches have already drawn scrutiny for their "black - box" characteristics, ViTs—while capable of some explainability through attention maps [2, 9]— still struggle to provide fully interpretable and trustworthy outputs. This complicates regulatory approval, clinical acceptance, and potential liability discussions. Furthermore, data-sharing inno- vations such as federated learning are only nascent and un tested in the context of disease detec- tion in laparoscopic video [4, 9]. Comparing ViT adoption in laparoscopic disease diagnosis to its success in endoscopic GI dis- ease tasks [4, 16], it becomes clear that pathologies with strong, localized visual signatures (e.g., certain GI lesions) align more easily with patch-based transformer reasoning. For heterogeneous, poorly demarcated diseases such as endometriosis in laparoscopy, future res earch will need to consider domain-adapted attention mechanisms or hybrid models that reintroduce local struc- tural biases [2, 8]. Moreover, workflow and tool recognition studies demonstrate the potential of training on larger, multi-institutional datasets [2, 8], a lesson directly transferable to the endome- triosis context if privacy and labeling barriers can be overcome. In summary, while there is strong comparative evidence that ViT models can achieve and sometimes exceed the performance of existing deep learning approaches in laparoscopy for tasks such as tool and workflow recognition [7, 8], there remains a marked lack of direct research on their application to automated disease diagnosis—and none at all on endometriosis detection in laparoscopic videos. The feasibility and ethical challenges identified here surpass those of previ- ous generations of AI models and must be systematically addressed before ViT-based diagnostics can be safely and successfully integrated into surgical practice. InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com

This study has several limitations. First, the literature search was restricted to English -lan- guage articles, potentially omitting relevant non-English research. Second, the lack of direct stud- ies on ViTs for endometriosis diagnosis necessitated extrapolation from adjacent tasks (e.g., tool detection), which may not fully represent disease -specific challenges. Third, the SWOT analysis, while systematic, remains qualitative and could benefit from quantitative validation through stakeholder surveys or multice ntric data. Finally, rapid advancements in transformer architec- tures may outdate some technical feasibility assessments.

Vision Transformers (ViTs) demonstrate potential for automating endometriosis diagnosis in lap- aroscopic videos, leveraging their ability to model global spatial relationships. However, this ap- plication remains understudied compared to tool or workflow recognition. Key barriers include data scarcity, computational costs, and unresolved ethical concerns. Future work must prioritize curated datasets, hybrid architectures for efficiency, and rigorous clinical validation to bridge the gap between technical promise and real-world utility. Abbreviations ViT: Vision Transformer; CNN: Convolutional Neural Network; RNN: Recurrent Neural Network; SWOT: Strengths, Weaknesses, Opportunities, Threats; GI: Gastrointestinal. Ethical approval This study analyzed published literature and did not involve human participants, primary data collection, or patient interactions. As a text -based review, it required no institutional ethics ap- proval or compliance with declarations such as the Helsinki Code. Availability of data and materials Please contact the corresponding author if you would like access to the datasets used and/or analyzed during this study. Funding This research was not funded or supported by any organizations. Authors’ Contribution P.H.: Conceptualization, literature review, manuscript drafting. AA.: Data curation, table/figure design, ethical analysis. F.R.: Methodology, literature screening, SWOT framework development. K.GH.: Abstract/keywords, limitations/conclusion drafting. H.GH.: References formatting, technical validation, final editing. Acknowledgment Not applicable. Consent for publication The authors provided their consent for the publication of the study results. Competing interests The authors declare no competing interests. InfoScience Trends ǁ (2025) NO 05; VOL 02: 11-22 ǁ ǁ https://doi.org/10.61186/ist.202502.05.02ǁ InfoPub www.isjtrend.com

[1]. Maheux-Lacroix S, Belanger M, Pinard L, Lemyre M, Laberge P, Boutin A. Diagnostic Accuracy of Intraoperative Tools for Detecting Endometriosis: A Systematic Review and Meta-analysis. Journal of minimally invasive gynecology. 2020;27(2):433-40.e1. https://doi.org/10.1016/j.jmig.2019.11.010 [2]. Kondo S. Lapformer: surgical tool detection in laparoscopic surgical video using transformer architecture. Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization. 2021;9(3):302-7. https://doi.org/10.1080/21681163.2020.1835550 [3]. Thakur GK, Thakur A, Kulkarni S, Khan N, Khan S. Deep learning approaches for medical image analysis and diagnosis. Cureus. 2024;16(5). https://doi.org/10.7759/cureus.59507 [4]. Yadav S, Aparna P, editors. Performance Comparison of Transformers and Convolutional Neural Networks (CNNs) Based Architecture on Endoscopy Images. 2024 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT); 2024: IEEE. https://doi.org/10.1109/CONECCT62155.2024.10677209 [5]. Alijani S, Fayyad J, Najjaran H. Vision transformers in domain adaptation and domain generalization: a study of robustness. Neural Computing and Applications. 2024;36(29):17979-8007. https://doi.org/10.1007/s00521- 024-10353-5 [6]. Gratton SM, Choudhry AJ, Vilos GA, Vilos A, Baier K, Holubeshen S, et al. Diagnosis of Endometriosis at Laparoscopy: A Validation Study Comparing Surgeon Visualization with Histologic Findings. Journal of obstetrics and gynaecology Canada : JOGC = Journal d'obstetrique et gynecologie du Canada : JOGC. 2022;44(2):135-41. https://doi.org/10.1016/j.jogc.2021.08.013 [7]. El Moaqet H, Janini R, Abdulbaki Alshirbaji T, Aldeen Jalal N, Möller K, editors. Using Vision Transformers for Classifying Surgical Tools in Computer Aided Surgeries. Current Directions in Biomedical Engineering; 2024: De Gruyter. https://doi.org/10.1515/cdbme-2024-2056 [8]. Zhang B, Abbing J, Ghanem A, Fer D, Barker J, Abukhalil R, et al. Towards accurate surgical workflow recognition with convolutional networks and transformers. Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization. 2022;10(4):349-56. https://doi.org/10.1080/21681163.2021.2002191 [9]. Abiyev RH, Altabel MZ, Darwish M, Helwan A. A Multimodal Transformer Model for Recognition of Images from Complex Laparoscopic Surgical Videos. Diagnostics. 2024;14(7):681. https://doi.org/10.3390/diagnostics14070681 [10]. Puyt RW, Lie FB, Wilderom CPM. The origins of SWOT analysis. Long Range Planning. 2023;56(3):102304. https://doi.org/10.1016/j.lrp.2023.102304 [11]. Rony MKK, Akter K, Debnath M, Rahman MM, Johra Ft, Akter F, et al. Strengths, weaknesses, opportunities and threats (SWOT) analysis of artificial intelligence adoption in nursing care. Journal of Medicine, Surgery, and Public Health. 2024;3:100113. https://doi.org/10.1016/j.glmedi.2024.100113 [12]. Nasirihaghighi S, Ghamsarian N, Husslein H, Schoeffmann K, editors. Event Recognition in Laparoscopic Gynecology Videos with Hybrid Transformers. International Conference on Multimedia Modeling; 2024: Springer. https://doi.org/10.1007/978-3-031-56435-2_7 [13]. Cao X, Guan H. Bidirectional transformer with sparse attention for gastrointestinal disease recognition. In2023 4th International Conference on Big Data, Artificial Intelligence and Internet of Things Engineering (ICBAIE) 2023 Aug 25 (pp. 357-361). IEEE. https://doi.org/10.1109/ICBAIE59714.2023.10281350 [14]. Arabian H, Abdulbaki Alshirbaji T, Jalal NA, Krueger-Ziolek S, Moeller K. P-CSEM: An Attention Module for Improved Laparoscopic Surgical Tool Detection. Sensors. 2023;23(16):7257. https://doi.org/10.3390/s23167257 [15]. Bajraktari F, Pott PP, editors. Multi-view surgical phase recognition during laparoscopic cholecystectomy. Current Directions in Biomedical Engineering; 2024: De Gruyter. https://doi.org/10.1515/cdbme-2024-2011 [16]. Demir KC, Schieber H, Weise T, Roth D, May M, Maier A, et al. Deep learning in surgical workflow analysis: a review of phase and step recognition. IEEE Journal of Biomedical and Health Informatics. 2023;27(11):5405- 17. https://doi.org/10.1109/jbhi.2023.3311628 Publisher’s Note © 2025 The Author(s). Published by InfoPub. Publisher homepage: https://infopub.ir/ Disclaimer: The views, opinions, and data presented in this article are solely those of the author(s) and do not necessarily reflect the official policy or position of InfoScience Trends or its editorial team. InfoScience Trends and its editors disclaim any liability for errors, consequences, or damages arising from the use of information contained in this publication.