Multimodal deep learning with anatomically constrained attention for screening MRI-detectable TMJ abnormalities from panoramic images - Takeaways - MDSpire

Multimodal deep learning with anatomically constrained attention for screening MRI-detectable TMJ abnormalities from panoramic images

  • By

  • Hyo-Jung Jung

  • Dayun Ju

  • Chanyoung Kim

  • Seong Jae Hwang

  • Chena Lee

  • Younjung Park

  • January 23, 2026

  • 0 min

Share

  • 1

    Early diagnosis of temporomandibular disorders (TMD) is difficult, especially for intra-articular TMJ abnormalities confirmed by MRI.

  • 2

    The study developed a deep learning framework using paired TMJ panoramic radiographs and clinical metadata for improved diagnosis.

  • 3

    The best-performing model achieved an area under the curve of 0.86, effectively classifying MRI-negative and -positive cases.

  • 4

    Gradient-weighted Class Activation Mapping confirmed the model's focus on condylar regions, highlighting its interpretability.

  • 5

    The proposed workflow aids in triaging TMJ patients for MRI referrals, facilitating early detection and timely interventions.

Original Source(s)

Multimodal deep learning with anatomically constrained attention for screening MRI-detectable TMJ abnormalities from panoramic images

  • by Hyo-Jung Jung, Dayun Ju, Chanyoung Kim, Seong Jae Hwang, Chena Lee, Younjung Park

  • January 23, 2026

Related Content

Evaluation of large language models for diagnostic impression generation from brain MRI report findings: a multicenter benchmark and reader study

Who Decides When a Doctor Steps Back?

Cognitive decline, credentialing, and why “I’ll know” often isn’t enough.

Magnetic resonance imaging in the diagnosis of trigeminal neuralgia: a systematic review of the imaging protocol and diagnostic accuracy