Novel Architecture of RPA In Oral Cancer Lesion Detection

Revana Magdy; Joy Naoum; Ali Hamdi

cs.CVMar 11, 2026

Novel Architecture of RPA In Oral Cancer Lesion Detection

Authors: Revana Magdy, Joy Naoum, Ali Hamdi

Organizations: MSA University, Giza, Egypt

Abstract

Accurate and early detection of oral cancer lesions is crucial for effective diagnosis and treatment. This study evaluates two RPA implementations, OC-RPAv1 and OC-RPAv2, using a test set of 31 images. OC-RPAv1 processes one image per prediction in an average of 0.29 seconds, while OCRPAv2 employs a Singleton design pattern and batch processing, reducing prediction time to just 0.06 seconds per image. This represents a 60-100x efficiency improvement over standard RPA methods, showcasing that design patterns and batch processing can enhance scalability and reduce costs in oral cancer detection

Figures & tables

Figure 1: The OC RPA workflow illustrates two parallel pipelines that converge at the same CNN model. The first pipeline processes a single image through the CNN model, while the second applies design patterns—specifically, the Singleton and Batch Strategy patterns—to process a batch of images.

Concept	General Definition (Nividous, 2025)	Application in This Project
RPA Purpose	Automates structured, rule-based processes without cognitive understanding.	UiPath Studio and Automation Anywhere automate the inference workflow—loading medical images, executing the trained CNN model, and logging prediction time.
AI Purpose	Mimics human intelligence through learning and decision-making from data.	The EfficientNetV2-B1 CNN model performs image-based oral cancer detection by learning visual patterns from training data.
Integration Point	RPA orchestrates process flow, while AI provides decision logic.	RPA platforms execute automation pipelines, while the CNN provides intelligent classification. Together, they form an Intelligent Automation (IA) system.
Task Nature	RPA: deterministic workflows; AI: analytical and adaptive reasoning.	RPA automates structured steps (e.g., file handling, interface control), while AI handles complex visual recognition and classification.
Error Handling	RPA is limited to predefined conditions and structured input.	RPA alone cannot interpret medical images; the AI model ensures robust predictions even with visual variability.
Performance Focus	RPA optimizes speed and consistency; AI optimizes accuracy and adaptability.	The project measures prediction time across Python, UiPath, and Automation Anywhere, balancing efficiency (RPA) and intelligence (AI).

Table 1: Mapping of RPA and AI Concepts to the Proposed Oral Cancer Detection Workflow

Panel of Figure 2 — Figure 2: Sample images from the dataset, showing different oral cancer lesion types (from , [ 11 ] ).

Metric	UiPath	Automation Anywhere	OC-RPA_V1	OC-RPA_V2
Folder Time (31 Images)	80 s	75 s	8.65 s	1.96 s
Avg Time per Image	2.58 s	2.42 s	0.28 s	0.06 s

Table 2: Execution time comparison among RPA and optimized Python-based implementations for oral cancer lesion prediction.

Panel of Figure 3 — Figure 3: Accessibility–Performance trade-off among UiPath, Automation Anywhere, OC-RPA v1, and OC-RPA v2. Dashed line denotes the Pareto frontier of optimal configurations.

Explore similar work

Robust Lightweight Deep Learning Models for Oral Cancer Screening
Aug 21, 2026Siddhant Bharadwaj, Aakash Shedsale, Tejashree Subramanya +6Medical Image Classification Lightweight CNNs
Indian Institute of Science · KLES’ Institute of Dental Sciences · Indian Institue of Science +1
RAPTOR+: A Visually Grounded Vision-Language Framework to Improve Clinical Trust and Auditability in Automated Cancer Referral Processing
May 25, 2026Sofiat Abioye, Ufaq Khan, Shazad Ashraf +4Document Understanding Healthcare
Birmingham City University, Birmingham, England, UK · Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), Abu Dhabi, UAE · University Hospitals Birmingham NHS Foundation Trust, Birmingham, England, UK +1
Detection and Classification of (Pre)Cancerous Cells in Pap Smears: An Ensemble Strategy for the RIVA Cervical Cytology Challenge
Mar 24, 2026Lautaro Kogan, María Victoria RíosComputational Pathology Class-Imbalanced Learning
Universidad de San Andrés, Buenos Aires, Argentina

Aug 21, 2026J/K move · Enter open · S save

Robust Lightweight Deep Learning Models for Oral Cancer Screening

Siddhant Bharadwaj, Aakash Shedsale, Tejashree Subramanya, Mohd. Azfar+5

Indian Institute of Science · KLES’ Institute of Dental Sciences · Indian Institue of Science+1

Medical Image Classification Lightweight CNNs Edge ML Hybrid CNN-Transformer Architectures Oncology

Open paper AskPDF arXiv

Oral cancer is a leading cause of mortality in low-to-middle-income countries, where a shortage of specialists delays diagnosis. While point-of-care screening via smartphones offers a scalable solution, developing robust AI for resource-constrained settings poses significant challenges, including class imbalance in training data, variable data quality, and computational constraints on edge devices. In this paper, we present the optimisation of lightweight deep learning models for smartphone-based oral cancer screening. Using a diverse, multi-centre retrospective dataset of approximately 30,000 images acquired over a decade, we systematically evaluate state-of-the-art convolutional, transformer, and hybrid architectures. Through rigorous pipeline ablation, we demonstrate that directly optimising hybrid architectures for the edge strictly outperforms computationally heavy paradigms, such as large models or knowledge distillation. Furthermore, interpretability analysis and simulated noise-stress tests revealed that the system anchors on clinical features and remains robust to unstructured sensor noise, despite vulnerabilities to impulse bit errors. In the held-out test set, our optimised MobileViTv2 models achieved an average sensitivity of 83.2

\pm

1.5% and an average specificity of 86.0

\pm

0.8%, with the best model exhibiting 87.4% sensitivity, 86.5% specificity, and a critical negative predictive value of 97.2% with reference to specialist labels. These results confirm that with targeted architectural selection and streamlined optimisation, interpretable and robust lightweight AI models exhibit high potential for edge deployment to enable automated triage in primary care settings.

Abstract

Explore similar work

Robust Lightweight Deep Learning Models for Oral Cancer Screening

RAPTOR+: A Visually Grounded Vision-Language Framework to Improve Clinical Trust and Auditability in Automated Cancer Referral Processing

Detection and Classification of (Pre)Cancerous Cells in Pap Smears: An Ensemble Strategy for the RIVA Cervical Cytology Challenge