🚀 AI vs. Experts in ANA Pattern Interpretation: Who Does Better?
Our recent study, published in Clinica Chimica Acta [1], compared an hierarchical stacked CNNs based software (akironNeo (propertitary 😒 )) with two experienced immunologists for interpreting ANA HEp-2 IIF assay patterns using 2,671 consecutive serum samples.
Key Findings:
✅ Good Agreement: The CNN showed good agreement (κ > 0.6) with human experts in classifying positive/negative results.
✅ Pattern Classification: Moderate to good agreement for pattern classification according to ICAP competent level, with some patterns like AC-3 Centromere showing very high agreement.
🔍 The CNN demonstrated cautious behavior at low fluorescence intensities by labeling some samples as indeterminate, potentially reducing false positives in routine diagnostics.
Our study shows theusefullness of CNN to facilitate diagnostic workflows by reducing subjectivity and improving standardization. However, challenges remain, such as developing standardized models that cover the ICAP expert level and ensuring high-quality annotated training data.
📄 Full paper: Schmidt, J., Weiß, S., Blessing, F., Blessing, J., Schierack, P., Rödiger, S., Hiemann, R., Roggenbuck, D., 2026. Comparison of manual with artificial intelligence-aided interpretation of ANA HEp-2 IIF assay patterns in a clinical diagnostics lab. Clinica Chimica Acta 584, 120881. doi.org/10.1016/j.cca.2026.1...
Albeit the technology is not that new, CNN are commonly used [5], we used a failry large data set, which is going to be published with some additional information. They have indeed advantages but, I still see a need for other approaches [2-4].
Current work
We are also working on an open source (GPL3) tool that offers a *Combination of Image Processing and Machine Learning as a novel method for foci detection. *
The tool is called Nucdetect github.com/SilMon/NucDetect. We hope to finish it this year.
Further Reading
[1] Schmidt, J., Weiß, S., Blessing, F., Blessing, J., Schierack, P., Rödiger, S., Hiemann, R., Roggenbuck, D., 2026. Comparison of manual with artificial intelligence-aided interpretation of ANA HEp-2 IIF assay patterns in a clinical diagnostics lab. Clinica Chimica Acta 584, 120881. doi.org/10.1016/j.cca.2026.1...
[2] Brauckhoff, T., Kieffer, C., Rödiger, S., 2024. biopixR: Extracting Insights from Biological Images. Journal of Open Source Software 9, 7074. doi.org/10.21105/joss.07074
[3] Brauckhoff, T., Rublack, J., Rödiger, S., 2025. Exploring Image Analysis in R: Applications and Advancements. The R Journal 17, 212–260. doi.org/10.32614/RJ-2025-030
[4] Schneider, J., Weiss, R., Ruhe, M., Jung, T., Roggenbuck, D., Stohwasser, R., Schierack, P., Rödiger, S., 2019. Open source bioimage informatics tools for the analysis of DNA damage and associated biomarkers. Journal of Laboratory and Precision Medicine 4, 1–27. doi.org/10.21037/jlpm.2019.0...
[5] Weiss, R., Karimijafarbigloo, S., Roggenbuck, D., Rödiger, S., 2022. Applications of Neural Networks in Biomedical Data Analysis. Biomedicines 10, 1469. doi.org/10.3390/biomedicines...
#ANA #AutoimmuneDiseases #AIinMedicine #Diagnostics #MachineLearning #CNN
