Skip to Content
BiologyBiology
  • Article
  • Open Access

23 June 2026

17 Pages

An Interpretable Deep Learning System for Fine-Grained Classification and Longitudinal Tracking of Neonatal Auricular Deformities

,
,
,
,
,
and
1
School of Medical Technology and Information Engineering, Zhejiang Chinese Medical University, Hangzhou 310053, China
2
Department of Otorhinolaryngology-Head and Neck Surgery, Xinhua Hospital, Shanghai Jiaotong University School of Medicine, Shanghai 200092, China
3
The Second School of Clinical Medicine, Zhejiang Chinese Medical University, Hangzhou 310053, China
*
Authors to whom correspondence should be addressed.

Simple Summary

At present, the clinical assessment of the auricle morphology of neonates only relies on subjective decisions, which often leads to neonates missing the short time window for early non-surgical treatment. Furthermore, there is currently no objective method to track the effectiveness of its treatment. To solve these problems, we developed a deep learning based decision-making and evaluation system to automatically diagnose and track ear conditions from basic photographs. The results show that our system can accurately screen for abnormal ears and identify six specific types of ear malformations. In addition, the deformity degree score can effectively assess the improvement before and after treatment. In conclusion, an easy-to-use web tool based on this system was developed to assist clinicians in making rapid diagnoses and tracking treatment outcomes, thereby improving clinical decision-making and patient outcomes.

Abstract

Early non-invasive correction of neonatal auricular deformities is highly dependent on timely and precise diagnosis. However, clinical practice is often compromised by the subjectivity of visual assessments and the lack of objective tracking metrics, which frequently leads to missed optimal treatment windows. To address these challenges, we developed an interpretable deep learning-based diagnostic system for the automated screening and fine-grained classification of these deformities. Methodologically, a large-scale, multi-source dataset (n = 4644) was curated to support model training. The system pairs an automated object detector (YOLOv11) for background-reduced region-of-interest isolation with a cascaded classification pipeline optimized via ConvNeXt-Tiny. Crucially, we introduced a supervised contrastive learning module to project high-dimensional morphological features into a continuous severity score, enabling quantitative longitudinal tracking of therapeutic efficacy. To evaluate generalization and robustness, the framework underwent rigorous evaluation across three independent real-world cohorts and one controlled synthetic stress test. The system achieved 88.2% accuracy (Area Under the Curve (AUC): 0.949) in binary screening and 87.4% accuracy (macro-AUC: 0.976) in multi-class subtyping on the internal baseline. To enhance interpretability and build clinical trust, Gradient-weighted Class Activation Mapping (Grad-CAM) was utilized to explore the spatial distribution of the model’s attention, which frequently aligned with key anatomical landmarks. Furthermore, the learned severity scores robustly quantified post-intervention improvements (p = 0.0004), effectively capturing subtle anatomical normalization. While validation for rare subtypes remains underpowered, and the severity score currently functions mainly as a learned morphological similarity index requiring future clinical calibration, this study ultimately provides an objective and standardized web-based tool to facilitate the early intervention and precision management of neonatal auricular anomalies.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.