A BAYESIAN FRAMEWORK FOR UNCERTAINTY QUANTIFICATION IN AI-DRIVEN DIAGNOSTIC TOOLS FOR RARE AND UNDERREPRESENTED CLINICAL PHENOTYPES
Keywords:
Bayesian inference, uncertainty quantification, AI diagnostics, rare diseases, clinical phenotypes, deep learning, probabilistic modelingAbstract
Rare and underrepresented clinical phenotypes pose significant challenges for artificial intelligence (AI)-based diagnostic systems due to limited data and inherent variability. This paper proposes a Bayesian framework for uncertainty quantification (UQ) in diagnostic models, allowing clinicians to assess prediction confidence and reduce diagnostic risk. The Bayesian approach provides a probabilistic perspective that naturally accommodates data scarcity and model ambiguity. We integrate this framework into a diagnostic pipeline using deep neural networks with Bayesian layers and test it on rare disease datasets. The results demonstrate improved interpretability and calibrated confidence estimates. This work underscores the necessity of incorporating UQ in AI diagnostics, especially for rare conditions where traditional models may be unreliable.
References
Ghassemi, M., Oakden-Rayner, L., Beam, A.L.: The false hope of current approaches to explainable artificial intelligence in health care. Lancet Digit. Health 4(11), e745–e750 (2022)
Gonepally, S., Amuda, K. K., Kumbum, P. K., Adari, V. K., & Chunduru, V. K. (2021). The evolution of software maintenance. Journal of Computer Science Applications and Information Technology, 6(1), 1–8. https://doi.org/10.15226/2474-9257/6/1/00150
Kendall, A., Gal, Y.: What uncertainties do we need in Bayesian deep learning for computer vision? Adv. Neural Inf. Process. Syst. 35, 5574–5584 (2022)
Jungo, A., Meier, R., Reyes, M.: Uncertainty-aware medical image segmentation with Bayesian neural networks. Med. Image Anal. 70, 102025 (2022)
Kong, W., Zhao, H., Fan, Y.: RareGenDB: A benchmark dataset for rare genetic disorder imaging tasks. Sci. Data 9, 442 (2022)
Amuda, K. K., Kumbum, P. K., Adari, V. K., Chunduru, V. K., & Gonepally, S. (2021). Performance evaluation of wireless sensor networks using the wireless power management method. Journal of Computer Science Applications and Information Technology, 6(1), 1–9. https://doi.org/10.15226/2474-9257/6/1/00151
Kompa, B., Snoek, J., Beam, A.L.: Second opinion needed: Communicating uncertainty reduces errors in AI-assisted diagnosis. NPJ Digit. Med. 5, 9 (2022)
Hu, Z., Xue, J., Zhang, X.: Bayesian neural network with Monte Carlo dropout for rare disease prediction. Appl. Intell. 52, 13811–13824 (2022)
Chunduru, V. K., Gonepally, S., Amuda, K. K., Kumbum, P. K., & Adari, V. K. (2022). Evaluation of human information processing: An overview for human-computer interaction using the EDAS method. SOJ Materials Science & Engineering, 9(1), 1–9.
Mo, S., Tian, C., Zhou, L.: UQNet: Deep uncertainty-aware learning for rare disease segmentation. Neural Comput. Appl. 34, 11091–11105 (2022)
Zhang, J., Wang, L., Yang, X.: Uncertainty-guided deep learning for early-stage rare disease detection. Comput. Biol. Med. 147, 105704 (2022)
Nguyen, M.T., Lee, J.H., Lee, S.G.: Bayesian hierarchical modeling of rare events in medical imaging using uncertainty-calibrated deep learning. Expert Syst. Appl. 203, 117497 (2022)
Ali, M., Rehman, S., Bashir, S.: Trustworthy AI in healthcare: A Bayesian ensemble framework for explainable rare disease diagnosis. J. Biomed. Inform. 131, 104083 (2022)
Chen, L., Wang, J., Xu, K.: Uncertainty-aware attention networks for rare disease image analysis. Pattern Recognit. 126, 108570 (2022)
Gonepally, S., Amuda, K. K., Kumbum, P. K., Adari, V. K., & Chunduru, V. K. (2022). Teaching software engineering by means of computer game development: Challenges and opportunities using the PROMETHEE method. SOJ Materials Science & Engineering, 9(1), 1–9.
Wang, Y., Li, J., Cheng, Y.: Quantifying aleatoric and epistemic uncertainties for rare phenotype detection using deep variational models. Med. Image Anal. 78, 102409 (2022)
Zhao, Y., He, L., Wang, B.: Rare lesion localization with uncertainty-aware generative learning. Comput. Med. Imaging Graph. 94, 101998 (2022)
Sun, X., Ma, J., Liu, P.: Calibrated diagnostic AI via Bayesian uncertainty: Addressing biases in rare clinical datasets. BMC Med. Inform. Decis. Mak. 22, 284 (2022)
Banerjee, I., Chen, M.C., Lungren, M.P.: Radiology report generation with uncertainty quantification for rare pathologies. IEEE Trans. Med. Imaging 41(1), 181–193 (2022)
Luo, H., Li, M., Zhou, X.: Robust Bayesian neural network for low-resource disease classification. Artif. Intell. Med. 125, 102227 (2022)
Patel, S., Singh, A., Sharma, R.: Epistemic uncertainty modeling in neural networks for rare dermatological condition diagnosis. Comput. Methods Programs Biomed. 220, 106817 (2022)
Downloads
Published
Issue
Section
License
Copyright (c) 2023 Bhavani Govindaraj, (Author)
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
