Dental-DPMNet: A Latent Disease Progression Modeling Network for Predicting Pulpitis and Periapical Disease Evolution Using Longitudinal Dental Imaging and Clinical Indicators

YingHan Li; TaoYu Zhu

doi:10.63313/JCSFT.9076

Authors

YingHan Li Henan University, Henan, 450001, China Author
TaoYu Zhu Johns Hopkins University, Baltimore, MD, 21218, USA Author

DOI:

https://doi.org/10.63313/JCSFT.9076

Keywords:

Dental disease progression, Pulpitis, Disease progression model, Neural ordinary differential equation, Longitudinal dental imaging, Multimodal learning

Abstract

Pulpal and periapical diseases exhibit a continuous and progressive pathological process driven by lesion evolution and inflammatory deterioration, yet most existing artificial intelligence approaches focus on static diagnosis and fail to model disease dynamics over time. In pharmaceutical research, Disease Progression Models (DPMs) have been widely adopted to characterize disease natural history using longitudinal biomarkers, enabling individualized prediction of disease trajectories. Inspired by this paradigm, this study proposes Dental-DPMNet, a latent disease progression modeling network for predicting the evolution of pulpitis and periapical diseases using longitudinal dental imaging and clinical indicators. Dental-DPMNet integrates cone-beam computed tomography and periapical radiographs with clinical examination records, including pain intensity, thermal test responses, and percussion sensitivity, to construct multimodal lesion evolution representations. A continuous latent disease state is introduced to capture the underlying severity of pulpal inflammation, and its temporal dynamics are explicitly modeled via a neural ordinary differential equation, allowing flexible handling of irregular follow-up intervals and patient-specific progression patterns. Multi-task decoders are employed to jointly reconstruct multimodal observations and estimate disease stage transition probabilities and future progression risks. Experiments on a retrospective longitudinal dental dataset demonstrate that Dental-DPMNet achieves an accuracy of 87.6% and an AUC of 0.91 in multi-stage prediction of pulpal and periapical disease progression. Compared with CNN-based, LSTM-based, and ODE-based baselines, the proposed model consistently shows superior performance. These results confirm the effectiveness of latent disease progression modeling for capturing lesion-driven evolution and supporting personalized endodontic decision-making.

References

[1] Sankaran K S. An improved multipath residual CNN-based classification approach for periapical disease prediction and diagnosis in dental radiography[J]. Neural Computing and Applications, 2022, 34(22): 20067-20082.

[2] Setzer F C, Shi K J, Zhang Z, et al. Artificial intelligence for the computer-aided detection of periapical lesions in cone-beam computed tomographic images[J]. Journal of endodontics, 2020, 46(7): 987-993.

[3] Ekert T, Krois J, Meinhold L, et al. Deep learning for the radiographic detection of apical lesions[J]. Journal of endodontics, 2019, 45(7): 917-922. e5.

[4] Wang, Yuyao. "Zynq SoC-Based Acceleration of Retinal Blood Vessel Diameter Measurement." Archives of Advanced Engineering Science (2025): 1-9.

[5] Lipton Z C, Kale D C, Elkan C, et al. Learning to diagnose with LSTM recurrent neural networks[J]. arXiv preprint arXiv:1511.03677, 2015.

[6] Pham T, Tran T, Phung D, et al. Deepcare: A deep dynamic memory model for predictive medicine[C]//Pacific-Asia conference on knowledge discovery and data mining. Cham: Springer International Publishing, 2016: 30-41.

[7] Jack C R, Knopman D S, Jagust W J, et al. Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade[J]. The Lancet Neurology, 2010, 9(1): 119-128.

[8] Young A L, Oxtoby N P, Daga P, et al. A data-driven model of biomarker changes in sporadic Alzheimer's disease[J]. Brain, 2014, 137(9): 2564-2577.

[9] Donohue M C, Jacqmin‐Gadda H, Le Goff M, et al. Estimating long‐term multivariate progression from short‐term data[J]. Alzheimer's & Dementia, 2014, 10: S400-S410.

[10] Chen R T Q, Rubanova Y, Bettencourt J, et al. Neural ordinary differential equations[J]. Advances in neural information processing systems, 2018, 31.

[11] Rubanova Y, Chen R T Q, Duvenaud D K. Latent ordinary differential equations for irregularly-sampled time series[J]. Advances in neural information processing systems, 2019, 32.

[12] De Brouwer E, Simm J, Arany A, et al. GRU-ODE-Bayes: Continuous modeling of sporadically-observed time series[J]. Advances in neural information processing systems, 2019, 32.

[13] Futoma J, Hariharan S, Heller K. Learning to detect sepsis with a multitask Gaussian process RNN classifier[C]//International conference on machine learning. PMLR, 2017: 1174-1182.

[14] Sun Q, Zhao X, Lin X. Design of a Hardware-Software Co-designed Real-Time Machine Learning System for Big Data Streams[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 265-271.

[15] Lin Y, Chen H, Chen X, et al. CEMG: Collaborative-Enhanced Multimodal Generative Recommendation[C]//International Conference on Multimedia Modeling. Singapore: Springer Nature Singapore, 2026: 508-521.

[16] Liang Z, Wei W, Zhang K, et al. Research on multi-hop inference optimization of llm based on mquake framework[J]. arXiv preprint arXiv:2509.04770, 2025.

[17] Lin, Ziyu, and Biliang Wang. "Adaptive load balancing algorithms for cloud computing distributed systems." IET Conference Proceedings CP952. Vol. 2025. No. 39. Stevenage, UK: The Institution of Engineering and Technology, 2025.