Artificial Intelligence-Based Distinction of Actinic Keratosis and Seborrheic Keratosis

Abstract

Actinic keratosis (AK) and seborrheic keratosis (SK) represent prevalent dermatological conditions with distinct clinical characteristics and potential health implications. This article investigates recent strides in dermatological diagnostics, centered on the development and application of artificial intelligence (AI) technology for discerning between AK and SK. The objective of this study is to develop and evaluate an artificial intelligence (AI) model capable of accurately distinguishing between stage one and stage two gastric carcinoma based on pathology slides. Employing a dataset of high-resolution images obtained from Kaggle.com, consisting of 1000 AK and 1000 SK images, a novel AI model was trained using cutting-edge deep learning methodologies. The dataset underwent meticulous partitioning into training, validation, and testing subsets to ensure robustness and generalizability. The AI model showcased exceptional proficiency in distinguishing AK from SK images, attaining notable levels of accuracy, precision, recall, specificity, F1-score, and area under the curve (AUC). Insights into the etiology and clinical ramifications of AK and SK were presented, emphasizing the critical significance of precise diagnosis and tailored therapeutic approaches. The integration of AI technology into dermatological practice holds considerable potential for enhancing diagnostic precision, refining treatment decisions, and elevating patient outcomes. This article underscores the transformative impact of AI in dermatology and the importance of collaborative efforts between clinicians, researchers, and technologists in advancing the realm of dermatological diagnosis and care.

Keywords: actinic keratosis; artificial intelligence (ai); dermatological imaging; lesion classification; seborrheic keratosis.

Figure 1. CNN Model Discerning Actinic Keratosis Images

The collection of images depicts discernible hallmarks specific to actinic keratosis, such as irregular borders, varying shades of coloration, and a rough, scaly texture. These distinctive features play a pivotal role in the detection and diagnosis process facilitated by the AI model. CNN: convolutional neural network; AI: artificial intelligence

Figure 2. CNN Model Identifying Various Images Representing Seborrheic Keratosis

The series of images showcases specific traits associated with seborrheic keratosis, such as well-defined borders, coloration ranging from tan to dark brown or black, and a waxy or stuck-on appearance. These characteristic features serve as essential indicators utilized by the AI model for accurate identification and diagnosis. CNN: convolutional neural network; AI: artificial intelligence

Figure 3. Precision-Recall Curve for the Model Distinguishing Actinic Keratosis and Seborrheic Keratosis

The graph illustrates the precision and recall capabilities of the neural network model at different confidence thresholds.

Figure 4. Confusion Matrix

Various metrics, including accuracy, precision, recall (sensitivity), specificity, and F-1 Score, were computed using data obtained from the confusion matrix.