MedLens

Please view the video above ☝️

Welcome!

My name is Omar Imrani and I am a grade 11 student attending Fraser Heights Secondary.

Please view my 1-minute explanation video above on MedLens a revolutionary solution to identify medical conditions.

Please continue navigating my presentation where will you find important data, the engineering process, code, design considerations, and other struggles I faced along the way.

➡️ Links:

Github Repository

Github Organization

Life Link - Web App (Please Visit on a Mobile Device)

My Story:

My inspiration for MedLens comes from a personal experience. A close family member of mine, my grandmother, was diagnosed with skin cancer, but it wasn’t discovered until it had already progressed. Early signs were overlooked because they didn’t seem serious at the time, and there wasn’t an easy way to quickly check or understand what was happening. Seeing how much earlier detection could have made a difference made me realize how important accessible and fast medical insight is.

Why Medical Detection?

Early detection is one of the most important factors in successfully treating many medical conditions, including skin cancer and tumors. However, many people face barriers such as limited access to healthcare, long wait times, and uncertainty about whether a condition is serious enough to seek medical attention. In many cases, individuals ignore early warning signs or delay getting checked due to lack of awareness or convenience. This delay can lead to worsening conditions that could have been treated more effectively if identified earlier. Additionally, people in remote areas or underserved communities often do not have immediate access to specialists, making early diagnosis even more difficult.

Why MedLens?

✅ AI-Powered Detection – Analyzes images to identify possible conditions, including skin cancer and tumors.

✅ Risk Level Assessment – Provides a clear indication of severity (low, medium, high).

✅ Instant Explanations – Helps users understand what the result means in simple terms.

✅ Recommended Actions – Suggests next steps, including whether to seek medical attention.

✅ Accessible-Design - Enables quick and easy use for anyone, anywhere.

Please view the slideshow above ☝️

This is a brief overview of the design process I followed to develop MedLens. For a more detailed explanation of the technical aspects and tools used, refer to the logbook below.

Logbook and Material for MedLens

How?

To develop MedLens, I first conducted background research on medical condition detection, focusing on how early identification of issues like skin cancer and tumors can improve outcomes. I used trusted sources such as medical websites, research articles, and public medical image datasets to understand how different conditions appear and how they are diagnosed.

Next, I designed and trained my own image classification model using a dataset of over 10,000 images. These images included a variety of medical conditions such as skin cancer, tumors, and more common injuries. I carefully selected and organized the data to ensure a balanced and diverse set of examples, which helped improve the model’s ability to recognize patterns accurately.

I then built the app using FlutterFlow, focusing on creating a simple and intuitive interface where users can take or upload a photo. For backend services, I used Firebase to handle image storage and data management. The trained model was integrated using Vertex AI, allowing the app to analyze images and return results such as possible conditions, risk levels, explanations, and recommended actions.

During testing, I evaluated the model using a separate set of images to check its accuracy and consistency. I tested different variables such as lighting, image clarity, and angles to ensure reliable performance. I also tested the full app workflow, including image upload, processing, and result display.

Through repeated testing and improvements, I refined both the model and the app to ensure MedLens provides fast, accessible, and helpful medical insights.

Please view the slideshow above ☝️

📊 Results & Analysis: How MedLens Works

MedLens is an AI-powered application designed to help users quickly analyze potential medical conditions, including skin cancer and tumors, using image-based detection. Through development and testing, the app demonstrated that AI can provide fast and accessible preliminary medical insights, helping users better understand possible health concerns.

🚀 Key Features & Results

✅ AI Image Analysis – The trained model, built using over 10,000 images, was able to recognize patterns across a wide range of medical conditions. It can analyze an uploaded image and return a possible condition within seconds.

✅ Risk Level Classification – Each result includes a clear risk level (low, medium, high), helping users understand the urgency of the situation.

✅ Instant Feedback – The app provides explanations in simple language, making complex medical information easier to understand.

✅ Recommended Actions – Users receive guidance on what to do next, such as monitoring the condition or seeking medical attention.

📱 How the MedLens App Works

The app is designed to be simple and efficient:

Users upload or capture an image of a medical concern

The image is securely stored and processed using Firebase

The AI model analyzes the image through Vertex AI

Results are displayed instantly, including condition type, risk level, and explanation

The interface, built using FlutterFlow, ensures that users can complete this process in just a few steps.

🌍 Accessibility & Global Use

MedLens is designed to be accessible to a wide range of users:

✅ Multilingual Support – The app can provide results in multiple languages, helping non-English speakers understand medical information clearly.

✅ Tourist-Friendly Use – Travelers in unfamiliar locations can quickly analyze a condition without needing to find a local clinic immediately, making it especially useful in emergencies or remote areas.

🚨 Emergency Support Integration

In high-risk situations, MedLens helps guide users toward immediate action:

✅ 911 / Emergency Guidance – If a condition is classified as high risk, the app recommends contacting emergency services or seeking urgent medical care.

✅ Clear Next Steps – Users are not left guessing; they are given direct, actionable advice based on the analysis

🏥 The Medical Profile – Personalized Emergency Assistance

One of MedLens's most valuable features is the Medical Profile, which gives emergency responders instant access to critical health details.

✅ Essential Health Information – Users can input allergies, chronic conditions, and medications.

✅ Emergency Contacts – Notifies selected contacts during a crisis.

✅ Special Needs & Accessibility – Details any conditions that responders should be aware of.

✅ Secure & Private – Data is protected and shared only when necessary.

🧪 Testing & Analysis

To evaluate performance, I tested the model using a separate set of images that were not included in training. These tests included variations in lighting, angle, and image quality to simulate real-world conditions. The model consistently provided relevant classifications and was able to distinguish between more serious conditions, such as tumors, and less severe issues like minor injuries.

Please view the slideshow above ☝️

So What?

The results of MedLens show that artificial intelligence can be used to provide fast and accessible medical insights, especially for early detection of conditions like skin cancer and tumors. From my testing, I learned that AI models trained on large datasets can recognize patterns and give useful preliminary assessments in seconds. This is important because early awareness can encourage people to seek medical help sooner, which can improve outcomes.

One key takeaway is that accessibility plays a major role in healthcare. MedLens helps reduce barriers by allowing users to analyze conditions from anywhere, without needing immediate access to a specialist. Features like clear explanations, risk levels, and multilingual support make the app easier to use for a wider range of people, including non-English speakers and travelers.

In addition, I demonstrated MedLens to multiple doctors and First Responders and received encouraging feedback. They appreciated the built-in translation and accessibility, and medical profile features, especially in time-critical situations.

I also sought reviews from several experienced software and data engineers working internationally at companies like Glovo, SaaSGlue, IBM, and Capgemini. They gave advice on technical improvements and praised the app’s clean code, secure data encryption, and overall architecture, emphasizing its privacy-first design, efficient backend, and suitability for real-world emergency deployment.

MedLens has the potential to help people recognize serious conditions earlier, take appropriate action, and reduce delays in seeking care, especially in situations where time and access to healthcare are critical.

🚀 What’s Next for MedLens?

MedLens has strong potential to improve early medical detection, and I plan to continue developing and expanding its capabilities. The next steps include:

✅ Improving Model Accuracy – Train the AI on larger and more diverse datasets to increase reliability and better detect a wider range of conditions.

✅ Expanding Condition Detection

✅ Real-Time Medical Integration – Explore connecting users with healthcare professionals for faster follow-up and support.

✅ AI Improvements & Personalization – Develop smarter responses tailored to individual users and their medical history.

✅ Improving Offline Functionality – Develop features for emergency use without internet access.

✅ Apply for a Patent

.

Acknowledgments

I acknowledge that our school is on the shared traditional territory of the Katzie, Semiahmoo, Kwantlen, and other Coast Salish Peoples.

I Would Like to Thank:

My Parents – For their support throughout this project.

I Would Like to Thank for Their Reviews and Software Development Help:

Mr. Abderrahman Bachiri Taoufiq – Senior Software Engineer, Glovo Barcelona

Mr. Issam Zoli – Senior Software Engineer, SaaSGlue

Mr. Tammame Azzaroual – Senior Data Engineer, IBM

Mr. Mohamed Benherref – Senior Java Tech Lead, Capgemini

➡️ Links:

Github Repository

Github Organization

Life Link - Web App (Please Visit on a Mobile Device)

References:

Amazon Web Services (AWS). (2025). Amazon SageMaker for medical imaging analysis. Retrieved February 12, 2026, from https://aws.amazon.com/sagemaker/

American Academy of Dermatology. (2024). Skin cancer types: Basal cell carcinoma, squamous cell carcinoma, and melanoma. Retrieved December 12, 2025, from https://www.aad.org/public/diseases/skin-cancer/types

American Cancer Society. (2025). Signs and symptoms of skin cancer. Retrieved January 12, 2026, from https://www.cancer.org/cancer/types/skin-cancer/symptoms-signs.html

Apple Developer Documentation. (2025). Integrating Core ML into your app for on-device medical imaging. Retrieved March 2, 2026, from https://developer.apple.com/documentation/coreml

British Journal of Dermatology. (2023). Evidence-based clinical practice guidelines for the management of patients with melanoma. Retrieved February 20, 2026, from https://academic.oup.com/bjd

Centers for Disease Control and Prevention. (2024). What is skin cancer? Retrieved December 28, 2025, from https://www.cdc.org/cancer/skin/basic_info/what-is-skin-cancer.htm

DermNet NZ. (2024). Skin cancer overview. Retrieved February 28, 2026, from https://dermnetnz.org/topics/skin-cancer

FlutterFlow. (2025). FlutterFlow documentation. Retrieved January 20, 2026, from https://docs.flutterflow.io/

Google Cloud. (2025). Vertex AI documentation. Retrieved March 30, 2026, from https://cloud.google.com/vertex-ai/docs

Google Firebase. (2025). Firebase documentation. Retrieved March 15, 2026, from https://firebase.google.com/docs

IEEE Xplore. (2024). Mobile application for early detection of skin cancer using deep learning. Retrieved December 22, 2025, from https://ieeexplore.ieee.org/

International Skin Imaging Collaboration. (2024). ISIC skin image dataset. Retrieved February 14, 2026, from https://www.isic-archive.com/

Journal of the American Medical Association (JAMA) Dermatology. (2024). Comparison of the Accuracy of Human Dermatologists vs Artificial Intelligence Algorithms. Retrieved January 30, 2026, from https://jamanetwork.com/journals/jamadermatology

Mayo Clinic. (2025). Skin cancer – Symptoms and causes. Retrieved March 22, 2026, from https://www.mayoclinic.org/diseases-conditions/skin-cancer/symptoms-causes/syc-20377605

National Cancer Institute. (2024). Skin cancer treatment (PDQ®)–patient version. Retrieved February 10, 2026, from https://www.cancer.gov/types/skin/patient/skin-treatment-pdq

National Library of Medicine. (2023). Artificial intelligence in medical imaging. Retrieved March 5, 2026, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7303818/

Nature Medicine. (2024). Clinical validation of artificial intelligence in dermatology. Retrieved March 10, 2026, from https://www.nature.com/nm/

Skin Cancer Foundation. (2025). Skin cancer facts & statistics. Retrieved January 15, 2026, from https://www.skincancer.org/skin-cancer-information/skin-cancer-facts/

Stanford University School of Medicine. (2017). Deep learning for skin cancer classification. Retrieved January 5, 2026, from https://cs.stanford.edu/people/esteva/nature/

TensorFlow. (2025). Image classification with convolutional neural networks. Retrieved March 18, 2026, from https://www.tensorflow.org/tutorials/images/classification

Tschandl, P., et al. (2025). The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions. Retrieved February 5, 2026, from https://pubmed.ncbi.nlm.nih.gov/30129051/

World Health Organization. (2024). Skin cancers. Retrieved December 15, 2025, from https://www.who.int/news-room/fact-sheets/detail/skin-cancers

Association for Computing Machinery (ACM). (2024). Ethical considerations in AI-based medical diagnostics. Retrieved January 15, 2026, from https://dl.acm.org/

Australian Institute of Health and Welfare. (2025). Skin cancer in Australia: Statistics and trends. Retrieved March 20, 2026, from https://www.aihw.gov.au/reports/cancer/skin-cancer-in-australia

BioMed Central (BMC) Medical Informatics. (2024). User interface design for mobile medical diagnostic apps. Retrieved February 12, 2026, from https://bmcmedinformdecismak.biomedcentral.com/

Cancer Research UK. (2024). Melonoma skin cancer statistics. Retrieved December 22, 2025, from https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/melanoma-skin-cancer

Cornell University (arXiv). (2025). EfficientNet: Rethinking model scaling for convolutional neural networks in medical imaging. Retrieved March 5, 2026, from https://arxiv.org/abs/1905.11946

European Academy of Dermatology and Venereology (EADV). (2025). Standards for teledermatology and digital imaging. Retrieved February 28, 2026, from https://eadv.org/

Food and Drug Administration (FDA). (2024). Artificial intelligence and machine learning (AI/ML) software as a medical device. Retrieved January 10, 2026, from https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device

Harvard Medical School. (2024). The role of AI in early cancer detection. Retrieved December 30, 2025, from https://hms.harvard.edu/news/ai-cancer-detection

Health Level Seven International (HL7). (2025). FHIR: Fast Healthcare Interoperability Resources for mobile health data exchange. Retrieved March 15, 2026, from https://www.hl7.org/fhir/

International Agency for Research on Cancer (IARC). (2024). World Cancer Report: Skin cancer prevention strategies. Retrieved January 25, 2026, from https://www.iarc.who.int/

Journal of Digital Imaging. (2023). Data augmentation techniques for skin lesion classification. Retrieved February 18, 2026, from https://link.springer.com/journal/10278

Journal of Medical Internet Research (JMIR). (2025). Patient trust in AI-driven dermatological screenings. Retrieved March 12, 2026, from https://www.jmir.org/

Kaggle. (2024). SIIM-ISIC Melanoma Classification Competition Dataset. Retrieved February 5, 2026, from https://www.kaggle.com/c/siim-isic-melanoma-classification

Lancet Digital Health. (2024). The future of diagnostic AI in low-resource settings. Retrieved January 5, 2026, from https://www.thelancet.com/journals/landig/home

Microsoft Azure. (2025). Custom Vision AI for healthcare image analysis. Retrieved March 28, 2026, from https://azure.microsoft.com/en-us/services/cognitive-services/custom-vision-service/

National Health Service (NHS). (2024). Diagnosing skin cancer: A guide for clinicians. Retrieved December 18, 2025, from https://www.nhs.uk/conditions/non-melanoma-skin-cancer/diagnosis/

Oxford Academic (Journal of Cybersecurity). (2024). Securing patient data in mobile diagnostic applications. Retrieved February 20, 2026, from https://academic.oup.com/cybersecurity

PubMed Central (PMC). (2024). Explainable AI (XAI) in medical imaging: A review. Retrieved March 8, 2026, from https://www.ncbi.nlm.nih.gov/pmc/

ScienceDirect. (2025). Deep learning architectures for skin lesion segmentation. Retrieved March 25, 2026, from https://www.sciencedirect.com/

U.S. Department of Health & Human Services. (2024). HIPAA compliance for mobile health app developers. Retrieved January 18, 2026, from https://www.hhs.gov/hipaa/for-professionals/special-topics/health-apps/index.html