Progressive Deep Learning for Automated Spheno-Occipital Synchondrosis Maturation Assessment

Omid Halimi Milani; Amanda Nikho; Marouane Tliba; Lauren Mills; Emadeldeen Hamdan; Ahmet Enis Cetin; Mohammed H. Elnagar

cs.CVApr 13, 2026

Progressive Deep Learning for Automated Spheno-Occipital Synchondrosis Maturation Assessment

Authors: Omid Halimi Milani, Amanda Nikho, Marouane Tliba, Lauren Mills, Emadeldeen Hamdan, Ahmet Enis Cetin, Mohammed H. Elnagar

Organizations: Department of Electrical and Computer Engineering, University of Illinois Chicago, Chicago, IL, USA · Department of Orthodontics, College of Dentistry, University of Illinois Chicago, Chicago, IL, USA · University of Orléans, Orléans, France

Abstract

Accurate assessment of spheno-occipital synchondrosis (SOS) maturation is a key indicator of craniofacial growth and a critical determinant for orthodontic and surgical timing. However, SOS staging from cone-beam CT (CBCT) relies on subtle, continuously evolving morphological cues, leading to high inter-observer variability and poor reproducibility, especially at transitional fusion stages. We frame SOS assessment as a fine-grained visual recognition problem and propose a progressive representation-learning framework that explicitly mirrors how expert clinicians reason about synchondral fusion: from coarse anatomical structure to increasingly subtle patterns of closure. Rather than training a full-capacity network end-to-end, we sequentially grow the model by activating deeper blocks over time, allowing early layers to first encode stable cranial base morphology before higher-level layers specialize in discriminating adjacent maturation stages. This yields a curriculum over network depth that aligns deep feature learning with the biological continuum of SOS fusion. Extensive experiments across convolutional and transformer-based architectures show that this expert-inspired training strategy produces more stable optimization and consistently higher accuracy than standard training, particularly for ambiguous intermediate stages. Importantly, these gains are achieved without changing network architectures or loss functions, demonstrating that training dynamics alone can substantially improve anatomical representation learning. The proposed framework establishes a principled link between expert dental intuition and deep visual representations, enabling robust, data-efficient SOS staging from CBCT and offering a general strategy for modeling other continuous biological processes in medical imaging.

Figures & tables

SOS Stage			Age Group
Stage	#	%	Group	#	%
Stage 1	159	22.0	Child (5 to 12)	281	38.9
Stage 2	92	12.7	Adolescent (13 to 19)	168	23.2
Stage 3	92	12.7	Adult (20 to 39)	116	16.0
Stage 4	125	17.3	Middle Age (40 to 59)	55	7.6
Stage 5	255	35.3	Senior Adult (60+)	9	1.2

Table 1: Expert-curated SOS CBCT dataset. We report 723 annotated scans across SOS maturation stages and subject age cohorts, highlighting class imbalance and broad age coverage.

Figure 3: Progressive-growing training. The backbone is expressed as an ordered sequence of blocks. At each stage, a single classification head is attached to the currently active depth and trained. Earlier heads are removed as deeper blocks are activated and training continues. At inference time, the final fully grown network is used with a single forward pass.

Model	$N$ (blocks)	Stage block counts $(n_{1},n_{2},n_{3},n_{4})$
ResNet–18	$8$	$(2,2,2,2)$
ResNet–34	$16$	$(4,4,4,4)$
ResNet–50	$16$	$(4,4,4,4)$
ResNet–101	$33$	$(8,8,8,9)$
ResNet–152	$50$	$(12,12,13,13)$

Table 2: Exact four-stage partitioning used in Scenario 1.

Model	$N$ (blocks)	Stage block counts $(n_{1},n_{2})$
ResNet–18	$8$	$(4,4)$
ResNet–34	$16$	$(8,8)$
ResNet–50	$16$	$(8,8)$
ResNet–101	$33$	$(16,17)$
ResNet–152	$50$	$(25,25)$

Table 3: Exact two-stage partitioning used in Scenario 2.

Figure 4: Illustration of progressive growing for Transformer encoders: at stage $k$ , only the first $k$ blocks are enabled and trained, while deeper blocks remain inactive until later stages.

Model	$L$	Stage sizes $(\|J_{1}\|,\|J_{2}\|)$
ViT-B/16	$12$	$(6,6)$
ViT-L/16	$24$	$(12,12)$

Table 4: Two-stage decomposition for ViT-B/16 and ViT-L/16.

Experiment	Model	Mode	Accuracy	Avg Metrics (Prec / Rec / F1)	Overall computation
320	ResNet-18	Entire	0.8603	0.8576 / 0.8603 / 0.8581	—
5, 5, 30, 280	ResNet-18	Progressive	0.8451	0.8451 / 0.8451 / 0.8440	89.3%
10, 290	ResNet-18	Progressive	0.8617	0.8599 / 0.8617 / 0.8595	96.8%
320	ResNet-101	Entire	0.8465	0.8433 / 0.8465 / 0.8441	—
5, 5, 30, 280	ResNet-101	Progressive	0.8600	0.8577 / 0.8589 / 0.8565	93.2%
400	ViT-B/16	Entire	0.4412	0.5165 / 0.4412 / 0.4325	—

Table 5: Comparison of entire-model and progressive-growing training across CNN and ViT backbones, including 32-patch ViT models with 50% patch overlap. Standard ResNet-18 training achieves an accuracy of 0.8603 after 320 epocs. Progressive training achieves an accuracy of 0.8617.

Explore similar work

Jun 15, 2026

Structure-aware Knowledge-guided Heterogeneous Mamba for Zygomaticomaxillary Suture Assessment

The Zygomaticomaxillary Suture is a key circummaxillary structure that connects the zygomatic bone and the maxilla, which serves as a primary site of resistance during maxillary advancement, and its maturation status directly influences the timing and efficacy of orthopedic interventions. However, accurate staging of ZMS maturation remains challenging due to subtle high-frequency transitions in suture lines and the global semantic ambiguity between adjacent stages. To address this, we present the first public ZMS dataset, comprising 3,790 ZMS images covering the entire age range from 4 to 24 years. Based on this dataset, we propose SKMamba, a Structure-aware and Knowledge-guided Mamba-based multi-modal framework for automated ZMS maturation assessment. SKMamba adopts a decoupled dual-path architecture that mimics the hierarchical diagnostic process used by experienced orthodontists. We first introduce an Implicit Edge Extractor (IEE), which leverages structural pre-training to reduce trabecular noise and accentuate sutural boundaries. Complementarily, a Cross-Modal Semantic Alignment (CSA) module is designed to incorporate anatomical descriptions from a large language model (LLM). This module helps align local morphological cues with global semantic descriptions while ensuring that objective morphological evidence remains the primary basis for decisions. Extensive experiments on our ZMS dataset demonstrate that SKMamba achieves state-of-the-art performance compared to existing methods. Code is available at https://github.com/galaxygxq1116/SKMamba.

Xiaoqi Guo, Birui Chen, Xinquan Yang +8

College of Computer Science and Software Engineering, Shenzhen University, Shenzhen, China · School of Artificial Intelligence, Shenzhen University, Shenzhen, China · Affiliated Stomatology Hospital of Kunming Medical University, Kunming, China +1

Medical Imaging Medical Image Classification

Ask PDF

Sep 11, 2026

LM-PCVMNet: Pediatric Cervical Vertebral Maturation Analysis with Deep Fusion of Landmarks and Metadata

Cervical vertebral maturation (CVM) assessment plays a pivotal role in orthodontic diagnosis and determining the optimal timing of treatment, especially for pediatric patients. In this paper, we propose LM-PCVMNet, a novel deep learning framework for automatic pediatric CVM staging. Specifically, our method integrates vertebral anatomical landmark information, heatmap-guided feature modulation, and metadata-informed similarity modeling into a unified learning framework. We introduce a heatmap-guided feature modulation module that enhances feature extraction by leveraging landmark-centered heatmaps to highlight morphologically relevant vertebral regions. A vertebral landmark-prompting block is designed to incorporate anatomical geometry into the representation learning process. Furthermore, we develop a learnable metadata supervised contrastive loss that adaptively modulates positive-pair similarity based on metadata similarity, enabling the model to learn more biologically consistent and discriminative features. To facilitate further research in pediatric orthodontic treatment, we additionally release PCVM+. It contains 1800 lateral cephalometric radiographs from real-world patients aged 3-15 years, with expert-annotated CVM stages, 13 vertebral anatomical landmarks, and corresponding metadata. We perform comprehensive experiments on two datasets, and the results show that our method achieves state-of-the-art performance, effectively improving landmark localization and classification accuracy over existing models. Code and dataset will be available at github.com/ybupengwang/LM-PCVMNet.

Peng Wang, Wanzhen Song, Anli Wang +3

College of Cryptology and Cyber Science, Nankai University, Tianjin, 300350, China · College of Engineering, Yanbian University, Yanji, 133002, Jilin, China · Haihe Lab of ITAI, Tianjin, 300350, China +1

Medical Imaging Medical Image Classification Medical Image Localization

Ask PDF

Apr 1, 2026

MATHENA: Mamba-based Architectural Tooth Hierarchical Estimator and Holistic Evaluation Network for Anatomy

Dental diagnosis from Orthopantomograms (OPGs) requires coordination of tooth detection, caries segmentation (CarSeg), anomaly detection (AD), and dental developmental staging (DDS). We propose Mamba-based Architectural Tooth Hierarchical Estimator and Holistic Evaluation Network for Anatomy (MATHENA), a unified framework leveraging Mamba's linear-complexity State Space Models (SSM) to address all four tasks. MATHENA integrates MATHE, a multi-resolution SSM-driven detector with four-directional Vision State Space (VSS) blocks for O(N) global context modeling, generating per-tooth crops. These crops are processed by HENA, a lightweight Mamba-UNet with a triple-head architecture and Global Context State Token (GCST). In the triple-head architecture, CarSeg is first trained as an upstream task to establish shared representations, which are then frozen and reused for downstream AD fine-tuning and DDS classification via linear probing, enabling stable, efficient learning. We also curate PARTHENON, a benchmark comprising 15,062 annotated instances from ten datasets. MATHENA achieves 93.78% mAP@50 in tooth detection, 90.11% Dice for CarSeg, 88.35% for AD, and 72.40% ACC for DDS.

Kyeonghun Kim, Jaehyung Park, Youngung Han +17

OUTTA, Seoul, South Korea · Seoul National University, Seoul, South Korea · Sangmyung University, Seoul, South Korea +4

Medical Imaging Medical Image Segmentation Medical Image Anomaly Detection Object Detection State Space Models

Ask PDF