Using AI to Assess Actinic Keratosis Stages with Imaging
Author Information
Author(s): Korecka Katarzyna, Slian Anna, Polańska Adriana, Dańczak-Pazdrowska Aleksandra, Żaba Ryszard, Czajkowska Joanna, Fleischer Alan
Primary Institution: Department of Dermatology, Poznan University of Medical Sciences
Hypothesis
Can machine learning algorithms improve the assessment of actinic keratosis stages using dermatoscopic and ultrasound images?
Conclusion
Machine learning algorithms can effectively assist in the staging of actinic keratosis, potentially improving clinical practice.
Supporting Evidence
- The accuracy of the dermatoscopic analysis using neural networks was 81%.
- Combining dermatoscopic and ultrasound scans achieved an accuracy of 79%.
- Machine learning can help predict the risk of progression in actinic keratosis.
- Statistical analysis showed significant differences in ultrasound features related to AK stages.
Takeaway
Doctors can use special computer programs to look at skin images and tell how serious a skin problem called actinic keratosis is.
Methodology
The study involved clinical, dermatoscopic, and ultrasound examinations of 54 patients, using machine learning algorithms for image analysis.
Potential Biases
Potential bias due to the small sample size and the reliance on expert classification.
Limitations
The study had a limited number of training examples, which may affect the accuracy of the machine learning models.
Participant Demographics
Patients aged 53 to 89, with a median age of 74, and 74% were males with Fitzpatrick I–II skin types.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website