The impact of physician experience on the development of machine learning in automation of general fecal examination using image processing
https://doi.org/10.33667/2078-5631-2025-22-65-69
Abstract
The creation of intelligent fecal image analysis software based on machine-learning is dependent on the quality of image annotation by experts. A pilot study revealed an extremely low degree of inter- and intra-expert agreement in recognizing morphological objects, as measured by the F1-score. Three types of discrepancies were identified: in quantity, presence, and category of objects. It is proposed to use a consensus-based approach to mitigate subjectivity and enhance the reliability of algorithms for automation of general fecal examination.
About the Authors
N. Yu. Chernysh
Almazov National Medical Research Centre
Russian Federation
Russian Federation
Chernysh Nataliia. Yu., PhD Med, associate professor at Dept of Laboratory Medicine with Clinic
St. Petersburg
V. S. Berestovskaya
Almazov National Medical Research Centre
Russian Federation
Russian Federation
Berestovskaya Victoria S., PhD Med, associate professor at Dept of Laboratory Medicine with Clinic
St. Petersburg
A. N. Tishko
Almazov National Medical Research Centre; All-Russian Center for Emergency and Radiation Medicine named after A.M. Nikiforov, Ministry of Emergency Situations of Russia
Russian Federation
Russian Federation
Tishko Anna N., PhD Med, Assistant at Dept of Laboratory Medicine with Clinic; clinical laboratory diagnostics physician
St. Petersburg
A. A. Rudneva
Institute of Medicine and Life Sciences, Immanuel Kant Baltic Federal University
Russian Federation
Russian Federation
Rudneva Anastasia A., head of Laboratory Diagnostics Cycle
Kaliningrad
T. V. Vavilova
Almazov National Medical Research Centre
Russian Federation
Russian Federation
Vavilova Tatyana V., DM Sci (habil.), professor, head of Dept of Laboratory Medicine with a Clinic
St. Petersburg
References
1. Евгина С. А., Гусев А. В., Шаманский М. Б., Годков М. А. Искусственный интеллект на пороге лаборатории. Лабораторная служба. 2022; 11 (2): 18–26. https://doi.org/10.17116/labs20221102118
2. Nkamgang O. T., Tchiotsop D., Fotsin H. B., Talla P. K., Dorr V. L., Wolf D. Automating the clinical stools exam using image processing integrated in an expert system. Informatics in Medicine Unlocked. 2019; 15: 100165. https://doi.org/10.1016/j.imu.2019.100165
3. Takahashi K., Yamamoto K., Kuchiba A., Shintani A., Koyama T. Hypothesis testing procedure for binary and multi-class F1 -scores in the paired design. Stat Med.2023; 42 (23): 4177–4192. doi: 10.1002/sim.9853
4. Никитин Е. Д. Моделирование индивидуального стиля разметки врача-рентгенолога для улучшения точности нейронных сетей // Digital Diagnostics. 2023;4 (1S): 99–101.
5. Alonso O. Challenges with Label Quality for Supervised Learning. Journal of Data and Information Quality (JDIQ), 2015; 6 (1): 1–3. https://doi.6.org/10.1145/2724721. Lekadir K., et al., FUTURE-AI Consortium FUTURE-AI: international consensus guideline for trustworthy and deployable artificial intelligence in healthcare. 2025. BMJ; 5:388: e081554. DOI: 10.1136/bmj-2024–081554.
Review
For citations:
Chernysh N.Yu., Berestovskaya V.S., Tishko A.N., Rudneva A.A., Vavilova T.V. The impact of physician experience on the development of machine learning in automation of general fecal examination using image processing. Medical alphabet. 2025;(22):65-69. (In Russ.) https://doi.org/10.33667/2078-5631-2025-22-65-69
Views:
79
Email this article 