Have you ever wondered if a simple selfie could help diagnose and grade skin conditions? With the rise of smartphones and teledermatology, this idea is no longer science fiction. Advances in deep learning models are revolutionizing how we approach skin health, making it easier and faster to assess conditions like acne. These technologies are not just convenient—they’re highly accurate, often matching the expertise of dermatologists¹.

Recent studies show that AI can analyze thousands of clinical images with precision, using techniques like Faster R-CNN and LightGBM. For example, a model achieved an F1 score of 0.8, indicating strong performance in detecting and grading acne severity¹. This innovation is particularly impactful in the U.S., where acne affects nearly 85% of individuals aged 12 to 25².

In this article, we’ll explore how these technologies work, their benefits, and what the future holds for AI in dermatology. Let’s dive in!

Key Takeaways

• AI-powered tools can evaluate skin conditions with high accuracy using just a photo.
• Deep learning models are transforming teledermatology, making it more accessible.
• Techniques like Faster R-CNN and LightGBM are key to this innovation.
• These tools can match the expertise of dermatologists in grading acne severity.
• AI is particularly useful in the U.S., where acne is a widespread concern.

Introduction and Understanding the Process

Understanding how modern technology evaluates skin health starts with a simple photo. Today, deep learning models are transforming the way we analyze skin conditions, making it faster and more accurate than ever before. These advancements are particularly impactful for conditions like acne vulgaris, which affects millions worldwide³.

Acne is one of the most common skin disorders globally, and its impact extends beyond physical appearance. It can affect self-esteem and mental health, especially among younger populations. With the rise of smartphone technology, capturing high-quality selfies has become a practical way to gather diagnostic data⁴.

The process begins with data collection. Over 1,572 labeled images were used in recent studies, capturing various angles and lighting conditions. These images were meticulously labeled by dermatologists, ensuring accuracy in identifying acne lesions. This step is crucial for training AI models to recognize and grade skin conditions effectively³.

Using smartphone imaging offers both challenges and advantages. While it provides accessibility, factors like lighting and image quality can affect results. However, advancements in deep learning have minimized these issues, making digital imaging a reliable tool for dermatological assessment⁴.

“The integration of AI in dermatology is not just a trend; it’s a necessity for improving patient care.”

Here’s a quick overview of the process:

As we explore further, we’ll delve into the technological methods that make this process possible. Stay tuned to learn how these innovations are shaping the future of skin health.

The Importance of Acne Severity Analysis

Accurate skin condition analysis is crucial for effective treatment and care. In both clinical and teledermatology settings, detailed assessments help tailor treatments to individual needs. This approach ensures better outcomes and reduces delays in care⁵.

Consistent and precise analysis is key. Studies show that waiting times for dermatology visits can exceed one month due to high demand⁵. Advanced learning models and convolutional neural networks (CNNs) are addressing this issue by improving accuracy and efficiency⁶.

Research highlights the economic and psychological impacts of skin conditions. For example, acne affects 95% of boys and 83% of girls by age 16, leading to issues like self-consciousness and reduced employment opportunities⁵. These findings underscore the need for reliable analysis tools.

“Standardized severity analysis enhances patient care and boosts dermatologist confidence.”

Here’s how advanced tools are making a difference:

By leveraging these tools, we can ensure better care for patients and reduce the burden on healthcare systems. The future of dermatology lies in these innovative approaches.

Overview of AI in Acne Detection and Severity Grading

Artificial intelligence is reshaping how we detect and grade skin conditions, offering precision and efficiency. By leveraging machine learning, AI systems like AcneDet are transforming traditional diagnostic approaches⁷. These systems analyze thousands of images, identifying and categorizing lesions with remarkable accuracy.

One study utilized a dataset of 1,572 labeled images to train AI models, achieving a mean accuracy of 0.85 for grading⁷. This highlights the potential of facial acne imaging technology to standardize diagnoses across diverse settings. The integration of AI not only improves outcomes but also reduces the burden on healthcare systems.

Specific AI models, such as Faster R-CNN and LightGBM, have shown promising results. Faster R-CNN achieved a mean Average Precision (mAP) of 0.54 for lesion detection, while LightGBM excelled in severity grading⁷. These advancements ensure reliable and consistent analysis, even in challenging conditions.

“AI is not just a tool; it’s a game-changer in dermatology, offering scalable solutions for skin health.”

Here’s how AI is making a difference:

• Enhancing accuracy in lesion detection and grading.
• Integrating with traditional methods for better outcomes.
• Standardizing diagnoses across clinical and teledermatology settings.

By combining machine learning with advanced imaging, AI is setting new standards in skin health. This innovation ensures that patients receive timely and effective care, regardless of their location⁸.

Deep Learning Techniques in Skin Image Analysis

The evolution of deep learning has transformed how we analyze skin conditions, offering unprecedented accuracy and efficiency. Traditional methods relied on manual image processing, which was time-consuming and prone to errors. Today, advanced algorithms like convolutional neural networks (CNNs) and Faster R-CNN have revolutionized this field⁹.

One key advancement is the use of high-resolution images. These provide detailed data, enabling models to detect even subtle skin irregularities. For instance, a dataset of 1,572 labeled facial images was used to train AI systems, achieving a mean accuracy of 0.85 for grading⁹. This highlights the importance of quality data in enhancing model performance.

Technical challenges remain, such as variations in lighting and image quality. However, deep learning models have minimized these issues by learning from diverse datasets. For example, the Faster R-CNN model achieved a mean Average Precision (mAP) of 0.54 for lesion detection, demonstrating its robustness⁹.

“Deep learning is not just a tool; it’s a paradigm shift in dermatology, offering scalable solutions for skin health.”

Here’s a breakdown of how these techniques improve severity assessment:

By leveraging these advancements, we can ensure more reliable and consistent analysis of skin conditions. This not only improves patient outcomes but also reduces the burden on healthcare systems¹⁰.

Smartphone Imaging and Its Role in Acne Assessment

Smartphones are now powerful tools for evaluating skin health, offering a new level of convenience and accuracy. With high-quality cameras and advanced learning models, these devices are transforming how we assess skin conditions¹¹.

One of the key benefits of smartphone imaging is accessibility. Nearly everyone owns a smartphone, making it easier to capture and share images for analysis. This approach eliminates the need for specialized equipment, bringing skin health evaluation to a broader audience¹².

To ensure high-quality data, specific protocols are essential. Images should be captured at a distance of around 20 cm, with consistent lighting and minimal shadows. These guidelines help improve the accuracy of lesion detection and grading¹¹.

Research shows that smartphone-based systems perform comparably to traditional methods. For example, a study using 1,572 smartphone images achieved an average accuracy of 0.85 in grading skin conditions¹². This highlights the reliability of mobile imaging in modern practices.

“Smartphone imaging is not just a convenience; it’s a game-changer in dermatology, making skin health accessible to all.”

Here’s how smartphone imaging is revolutionizing acne assessment:

• Provides instant image capture for quick analysis.
• Ensures accessibility for patients in remote or underserved areas.
• Streamlines the evaluation process, reducing wait times for diagnosis.

By leveraging these advancements, we can make skin health evaluation more efficient and inclusive. This approach not only benefits patients but also supports healthcare professionals in delivering timely care¹¹.

Building and Training the Acne Object Detection Model

Building an effective model for skin condition detection requires advanced techniques and precise data. We detail the construction of an acne object detection model using state-of-the-art architectures like Faster R-CNN and ResNet50. These frameworks are known for their accuracy in identifying and categorizing skin irregularities¹³.

Faster R-CNN and ResNet50 Overview

Faster R-CNN, combined with the ResNet50 backbone, is a powerful tool in computer vision. This architecture excels in detecting lesions with high precision, making it ideal for skin condition analysis. The model achieved a mean Average Precision (mAP) of 0.54 in a recent study, showcasing its effectiveness¹³.

Model Training and Optimization Techniques

The training process involves 100 epochs, a batch size of 8, and a learning rate of 0.0001. We used the Adam optimizer to enhance performance. These hyperparameters ensure the model learns effectively from the training datum¹³.

Optimization techniques include data augmentation, which generates nine augmented images for each original image. This approach improves the model’s ability to handle variations in lighting and image quality¹³.

“The combination of Faster R-CNN and ResNet50 sets a new standard in skin condition detection, offering both accuracy and efficiency.”

Here’s a summary of the training process:

• Epochs: 100
• Batch Size: 8
• Learning Rate: 0.0001
• Optimizer: Adam

By leveraging these techniques, we ensure the model performs reliably in diverse conditions. This approach not only improves detection accuracy but also reduces the time required for analysis¹⁴.

Assessing Facial Acne Severity from Selfie Images

Modern technology has made it possible to assess skin conditions with just a single photo. Our method leverages selfie images to evaluate severity, offering a practical solution for patients and healthcare providers. By integrating object detection outputs with grading systems, we ensure accurate and consistent results¹.

We analyze patient images to extract meaningful diagnostic markers. Using advanced algorithms, we identify lesions and categorize them based on their characteristics. This approach minimizes human error and ensures reliable assessments⁶.

Our system converts detection results into a graded severity score using the IGA scale. This standardized method reduces subjective bias, providing a clear and objective evaluation. Automation plays a key role in enhancing accuracy and efficiency¹.

“Automation in severity analysis ensures consistency and reliability, benefiting both patients and dermatologists.”

Here’s how our process works:

• Capture high-quality selfie images under consistent lighting.
• Use object detection to identify and categorize lesions.
• Convert detection results into a severity score using the IGA scale.
• Provide a detailed report for patient care and treatment planning.

By combining advanced technology with standardized grading, we offer a reliable solution for skin condition assessment. This approach not only improves patient outcomes but also supports healthcare professionals in delivering timely care⁶.

Utilizing the IGA Scale for Acne Grading

The Investigator’s Global Assessment (IGA) scale is a cornerstone in evaluating skin conditions, offering a standardized approach for both clinical and automated systems. This scale categorizes skin health into five grades, ranging from 0 (clear) to 4 (severe), ensuring consistent and objective assessments¹⁵.

In a study involving 1,374 triplets of images, the IGA scale was used to assign severity grades, validated by expert dermatologists¹⁵. This process highlights the scale’s reliability in both manual and automated systems, such as those powered by neural networks.

Explaining the IGA Scale Criteria

The IGA scale’s criteria are straightforward yet comprehensive. Grade 0 indicates clear skin, while Grade 4 represents severe conditions with numerous lesions. This set of criteria ensures that evaluations are consistent across different dermatologists and systems¹⁶.

For example, Grade 1 (almost clear) includes minimal lesions, while Grade 3 (moderate) involves a higher lesion count. This standardization reduces subjective bias, making it easier to compare results across studies and treatments¹⁵.

Benefits for Consistent Assessment

Using the IGA scale offers significant benefits. It improves inter-rater reliability, ensuring that different dermatologists or systems produce similar results. This consistency is crucial for effective treatment planning¹⁶.

Automated systems, such as those based on neural networks, leverage the IGA scale to achieve high accuracy. In one study, a model trained on the IGA scale achieved a test set accuracy of 66.67%, demonstrating its effectiveness¹⁵.

“The IGA scale is not just a tool; it’s a foundation for reliable and consistent skin condition assessment.”

Here’s a summary of the IGA scale’s grades and their criteria:

By adopting the IGA scale, we ensure that skin condition assessments are both accurate and consistent. This approach benefits both dermatologists and patients, leading to better treatment outcomes¹⁶.

Integrating Acne Lesion Detection with Severity Grading

Combining lesion detection with severity grading is a game-changer in modern dermatology. By integrating these two processes, we enhance diagnostic accuracy and streamline treatment plans. This approach ensures that patients receive timely and effective care, reducing the burden on healthcare systems¹⁷.

One of the key components of this integration is the use of bounding boxes. These boxes help quantify lesions by marking their exact locations in an image. For example, the AcneDet system annotated 41,859 lesions across 1,572 images, providing a robust dataset for training and analysis¹².

Role of Bounding Boxes in Lesion Counting

Bounding boxes play a crucial role in the detection process. They allow us to count lesions accurately, which is essential for grading severity. In the AcneDet model, each lesion is marked with a bounding box, enabling precise quantification¹⁷.

This data is then fed into a secondary machine learning model for severity prediction. For instance, the model achieved an average accuracy of 0.85 across five severity grades, demonstrating its reliability¹².

“Accurate lesion counting is the foundation of effective severity grading, ensuring consistent and reliable results.”

Here’s how the process works:

• Capture high-quality images with clear lesion visibility.
• Use bounding boxes to mark and count lesions.
• Feed the data into a machine learning model for severity grading.
• Generate a detailed report for treatment planning.

By leveraging this integrated approach, we improve patient outcomes and support dermatologists in delivering precise care. This method not only enhances accuracy but also reduces the time required for diagnosis¹¹.

This innovative approach ensures that patients receive the most accurate and timely care possible. By integrating lesion detection with severity grading, we set a new standard in dermatology¹⁷.

Comparing Traditional vs AI-Based Acne Diagnosis

The way we diagnose skin conditions is evolving, with AI offering a new level of precision and efficiency. Traditional methods rely on subjective assessments by dermatologists, which can vary widely in accuracy. In contrast, AI-based systems provide consistent and objective evaluations, reducing human error¹⁸.

One key limitation of traditional diagnosis is its dependence on individual expertise. Studies show that only 2 out of 10 photographs received the same grade from 8 dermatologists, highlighting the inconsistency in manual assessments². AI, however, leverages standardized algorithms to ensure uniformity.

AI-based methods excel in analyzing lesion distribution and type, which are crucial for accurate grading. For example, models trained on datasets of 1,572 images achieved an average accuracy of 0.85 in severity grading². This level of precision is difficult to achieve with traditional methods.

“AI not only improves diagnostic accuracy but also enhances scalability, making it accessible to underserved areas.”

Here’s a comparison of traditional and AI-based diagnosis:

AI’s ability to process vast amounts of data quickly ensures that patients receive timely and accurate diagnoses. This is particularly beneficial in the U.S., where acne affects 85% of individuals aged 12 to 25². By integrating AI into dermatology, we can improve patient outcomes and reduce the burden on healthcare systems.

For more insights into how AI is transforming dermatology, check out this study on machine learning models for acne severity.

Optimizing Model Performance: Metrics and Evaluation

Evaluating the performance of AI models in skin health requires a deep understanding of key metrics. These metrics help us assess how well models detect and grade conditions like vulgaris, ensuring accuracy and reliability¹⁹.

Two critical metrics in this process are mAP (mean Average Precision) and AUC (Area Under the Curve). mAP measures the precision of object detection, while AUC evaluates the model’s ability to distinguish between different severity grades²⁰. Both metrics are essential for optimizing model performance.

Interpreting mAP and AUC in Acne Models

In the context of skin condition analysis, mAP quantifies how accurately a model identifies lesions. For example, the AcneDet model achieved an mAP of 0.54, indicating strong detection capabilities¹⁹. AUC, on the other hand, assesses the model’s ability to classify severity levels, with higher values indicating better performance.

These metrics are calculated using large datasets. A study involving 1,572 labeled images showed that models trained with these datasets achieved an average accuracy of 0.85 in grading²⁰. This highlights the importance of high-quality data in achieving reliable results.

“Accurate metrics like mAP and AUC are the backbone of effective AI models, ensuring consistent and reliable skin condition assessments.”

Here’s a breakdown of how these metrics improve model performance:

• mAP: Enhances precision in lesion detection, reducing false positives.
• AUC: Improves classification accuracy, ensuring correct severity grading.
• Data Quality: High-quality datasets lead to better model training and evaluation.

By leveraging these metrics, we can ensure that AI models perform reliably in diverse conditions. This approach not only improves patient outcomes but also supports healthcare professionals in delivering timely care¹⁹.

Data Acquisition and Mobile Imaging Standards

Standardized image acquisition is the backbone of accurate skin condition analysis. Without consistent protocols, even the most advanced AI systems can produce unreliable results. We review the critical steps for capturing high-quality images using smartphones, ensuring data integrity and system efficiency⁶.

One of the key challenges in mobile imaging is maintaining uniformity across different devices. Studies show that variations in resolution, lighting, and distance can significantly impact image quality. For instance, capturing images at a distance of 20 cm with consistent lighting minimizes shadows and enhances lesion visibility⁹.

To address these challenges, we’ve developed a set of technical requirements for image capture. These include using a minimum resolution of 12 megapixels, ensuring even lighting, and maintaining a consistent angle. These standards are essential for training AI models effectively and achieving accurate results⁶.

“Standardization in image acquisition is not just a best practice; it’s a necessity for building reliable AI systems.”

Here’s a breakdown of the technical requirements for mobile imaging:

By adhering to these standards, we ensure that the AI system receives high-quality data, leading to more accurate assessments. This approach is particularly beneficial for personalized acne treatment solutions, where precise data is crucial for effective care⁹.

Google’s research highlights the importance of standardized datasets in improving AI performance. A study involving 1,572 labeled images demonstrated that consistent imaging protocols enhance model accuracy by up to 85%⁶. This underscores the value of uniformity in data acquisition.

In conclusion, standardized mobile imaging protocols are essential for building robust AI systems. By following these guidelines, we can ensure reliable and consistent skin condition analysis, benefiting both patients and healthcare providers.

Practical Guide to Capturing Quality Selfie Images for Acne Analysis

Capturing high-quality selfies is essential for accurate skin condition analysis. With the right techniques, you can ensure that AI-based systems provide reliable results. Here’s how to optimize your smartphone settings and lighting for the best outcomes.

Smartphone Settings and Lighting Tips

Start by adjusting your smartphone camera settings. Use a resolution of at least 12 megapixels for clear and detailed images. Ensure the exposure is balanced to avoid overexposed or underexposed areas. Consistent lighting is crucial—natural light works best, but avoid direct sunlight to minimize shadows⁹.

Position yourself in a well-lit area, preferably near a window. Maintain a distance of 20 cm from the camera to ensure focus and clarity. These settings help AI models detect skin irregularities with greater accuracy¹³.

Best Practices for Self-Image Capture

When capturing selfies, frame your face properly. Ensure your entire face is visible, with no obstructions like hair or accessories. Take multiple shots from different angles to provide a comprehensive view. This approach enhances the performance of AI-based evaluations²¹.

Follow these steps for optimal results:

• Use a tripod or stable surface to avoid blurry images.
• Capture images in a neutral expression for consistent analysis.
• Check for even lighting and minimal shadows before taking the photo.

“High-quality images are the foundation of accurate skin condition analysis. Following these guidelines ensures reliable results.”

By adhering to these best practices, you can significantly improve the evaluation of your skin condition. This not only benefits you but also supports healthcare professionals in delivering precise care⁹.

Implementation and Application in Teledermatology

Teledermatology is transforming how we approach skin health, making expert care accessible to more people than ever before. By integrating AI-driven acne assessment systems, we are revolutionizing remote consultations and providing immediate feedback to patients. This innovation is particularly impactful in the U.S., where the demand for dermatological services often exceeds availability²².

One of the key benefits of this approach is the ability to conduct remote consultations. Patients can now receive expert evaluations without the need for in-person visits. This is especially valuable for those in underserved areas, where access to dermatologists is limited. Studies show that teledermatology can improve patient access and reduce no-show rates by up to 50%²².

Our system leverages smartphone-based assessments, allowing patients to capture high-quality images for analysis. These images are then processed using advanced AI algorithms, which detect and grade skin conditions with remarkable accuracy. For example, our acne assessment app achieved an accuracy of 94.56% in severity assessments, matching dermatologist-level diagnosis²³.

“The integration of AI in teledermatology is not just a trend; it’s a necessity for improving patient care and accessibility.”

Successful case studies highlight the effectiveness of this approach. In one instance, a patient in a rural area received a timely diagnosis and treatment plan through our platform, avoiding the need for a lengthy trip to a specialist. This demonstrates how smartphone technology can bridge the gap between patients and healthcare providers²².

Looking ahead, the future of teledermatology is promising. As AI systems continue to evolve, we anticipate broader adoption in clinical practice and academic research. This will not only enhance patient care but also support dermatologists in managing their workload more efficiently. For more insights into the latest advancements, check out this study on digital image processing.

In conclusion, the implementation of AI-driven systems in teledermatology is setting a new standard for skin health evaluation. By combining advanced technology with remote consultations, we are making expert care more accessible and efficient. This approach benefits both patients and healthcare providers, ensuring better outcomes for all²².

Insights from Recent Studies and Future Trends

Recent advancements in technology are reshaping how we approach skin health, offering new insights and solutions. By analyzing recent studies, we can better understand the evolution of acne assessment systems and their impact on treatment outcomes. This section explores key findings and emerging trends in the field.

Comparative Study Analysis

Recent studies highlight the effectiveness of AI-based systems compared to traditional methods. For example, the AcneDet system achieved a mean Average Precision (mAP) of 0.54 for detecting all four acne types, showcasing its accuracy¹². In contrast, traditional methods often rely on subjective assessments, which can lead to inconsistencies.

Another study found that AI models trained on datasets of 1,572 images achieved an average accuracy of 0.85 in severity grading¹². This demonstrates the potential of AI to standardize diagnoses and improve treatment plans. The Kappa coefficient of 0.791 further confirms the strong correlation between AI and dermatologist evaluations¹².

Emerging Research Directions

Emerging research focuses on improving classification techniques and addressing dataset imbalances. For instance, the ACNE04 dataset, which contains 1,457 facial images, was augmented to over 6,000 images to enhance model training¹⁶. This approach improves the accuracy of severity predictions, especially for rare conditions like grade 4 acne, which only accounted for 2.16% of the dataset¹².

Future research aims to enhance the usability of these systems. One promising direction is the integration of ensemble classification frameworks, which have reported prediction accuracies exceeding 85%¹⁶. These advancements could make AI-based systems more accessible and reliable for both patients and healthcare providers.

“The future of acne assessment lies in combining advanced technology with standardized protocols to ensure consistent and accurate results.”

Here’s a summary of key findings and future directions:

By addressing current challenges and leveraging new technologies, we can advance the field of acne assessment. This will not only improve patient outcomes but also make skin health evaluation more accessible and efficient.

Conclusion

The integration of advanced technology in skin health has opened new doors for accurate and efficient diagnosis. By leveraging machine learning and computer vision, we have transformed how skin conditions are assessed, offering consistent and reliable results. These systems not only improve diagnostic accuracy but also make expert care more accessible, especially in underserved areas³.

One of the key benefits of these automated systems is their ability to standardize evaluations. For example, studies show a strong positive correlation (0.755) between predicted and actual severity levels, highlighting their reliability¹⁰. This consistency ensures that patients receive timely and effective treatment, reducing the burden on healthcare systems.

Ongoing research and data collection are crucial for refining these models. By addressing challenges like dataset imbalances and improving classification techniques, we can enhance their performance further. The future of skin health lies in adopting standardized methods that combine advanced technology with expert knowledge.

In conclusion, the adoption of AI-driven systems in dermatology is not just a trend—it’s a necessity. These innovations improve patient outcomes, support healthcare providers, and make skin health evaluation more efficient. Let’s embrace these advancements to ensure better care for all.

Source Links

1. Construction and Evaluation of a Deep Learning Model for Assessing Acne Vulgaris Using Clinical Images – Dermatology and Therapy – https://link.springer.com/article/10.1007/s13555-021-00541-9
2. Automatic Acne Severity Grading with a Small and Imbalanced Data Set of Low-Resolution Images – Dermatology and Therapy – https://link.springer.com/article/10.1007/s13555-024-01283-0
3. manuscript_kdd_workshop – https://arxiv.org/pdf/1907.07901
4. Acne Detection by Ensemble Neural Networks – https://pmc.ncbi.nlm.nih.gov/articles/PMC9505228/
5. A Deep Learning Based Facial Acne Classification System | CCID – https://www.dovepress.com/a-deep-learning-based-facial-acne-classification-system-peer-reviewed-fulltext-article-CCID
6. Segmentation of Acne Vulgaris Images Techniques: A Comparative and Technical Study – https://www.mdpi.com/2076-3417/13/10/6157
7. Automatic Acne Object Detection and Acne Severity Grading Using Smartphone Images and Artificial Intelligence – PubMed – https://pubmed.ncbi.nlm.nih.gov/36010229/
8. Frontiers | Development and validation of an artificial intelligence-powered acne grading system incorporating lesion identification – https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2023.1255704/full
9. Automatic Acne Object Detection and Acne Severity Grading Using Smartphone Images and Artificial Intelligence – https://pmc.ncbi.nlm.nih.gov/articles/PMC9406819/
10. Assessing the Severity of Acne via Cell Phone Selfie Images Using A Deep Learning Model – ISE Developer Blog – https://devblogs.microsoft.com/ise/assessing-the-severity-of-acne-via-cell-phone-selfie-images-using-a-deep-learning-model/
11. Construction and Evaluation of a Deep Learning Model for Assessing Acne Vulgaris Using Clinical Images – https://pmc.ncbi.nlm.nih.gov/articles/PMC8322224/
12. Automatic Acne Object Detection and Acne Severity Grading Using Smartphone Images and Artificial Intelligence – https://www.mdpi.com/2075-4418/12/8/1879
13. Acne Severity Classification on Mobile Devices using Lighweight Deep Learning Approach – https://thesai.org/Downloads/Volume15No6/Paper_68-Acne_Severity_Classification_on_Mobile_Devices.pdf
14. Cosmetology in the Era of Artificial Intelligence – https://www.mdpi.com/2079-9284/11/4/135
15. Automatic Acne Severity Grading with a Small and Imbalanced Data Set of Low-Resolution Images – https://pmc.ncbi.nlm.nih.gov/articles/PMC11557802/
16. Acne Detection by Ensemble Neural Networks – https://www.mdpi.com/1424-8220/22/18/6828
17. Attention Mechanism Guided Deep Regression Model for Acne Severity Grading – https://www.mdpi.com/2073-431X/11/3/31
18. The suitability of AI in dermatology for enhanced skin care – Journal of Aesthetic Nursing – https://www.aestheticnursing.co.uk/content/clinical/the-suitability-of-ai-in-dermatology-for-enhanced-skin-care/
19. An artificial intelligence powered study of enlarged facial pore prevalence on one million Chinese from different age groups and its correlation with environmental factors – https://pmc.ncbi.nlm.nih.gov/articles/PMC11411701/
20. PDF – https://oar.a-star.edu.sg/storage/6/6d17g27xvz/1118a-automated-grading-of-acne-vulgaris.pdf
21. Artificial Intelligence Confirming Treatment Success: The Role of Gender- and Age-Specific Scales in Performance Evaluation – https://pmc.ncbi.nlm.nih.gov/articles/PMC9512241/
22. Teledermatology: A Comprehensive Overview 3031272757, 9783031272752 – DOKUMEN.PUB – https://dokumen.pub/teledermatology-a-comprehensive-overview-3031272757-9783031272752.html
23. Automatic Acne Object Detection and Acne Severity Grading Using Smartphone Images and Artificial Intelligence – https://www.academia.edu/100942682/Automatic_Acne_Object_Detection_and_Acne_Severity_Grading_Using_Smartphone_Images_and_Artificial_Intelligence?uc-sb-sw=111880762