Thyroid Nodule Segmentation in Ultrasound Images with Conformal Uncertainty Estimation

Francesco Teodori; Pierpaolo Brutti; Umberto Ferraro Petrillo

cs.CVOct 9, 2026

Thyroid Nodule Segmentation in Ultrasound Images with Conformal Uncertainty Estimation

Authors: Francesco Teodori, Pierpaolo Brutti, Umberto Ferraro Petrillo

Organizations: Department of Statistical Sciences, Sapienza University of Rome, Rome, Italy.

Abstract

Thyroid nodules affect a large proportion of the general population. Although most are benign, a non-negligible fraction is malignant. Ultrasound imaging is the primary diagnostic modality, and accurate nodule segmentation is a prerequisite for CAD systems to classify malignancy and assess metastatic risk.\ In this work, we propose a thyroid nodule segmentation framework that integrates a DeepLabV3+, channel-attention modules and post-hoc conformal prediction, providing both accurate mask and pixel-level uncertainty estimates. We address thyroid ultrasound nodule segmentation as the first stage of a pipeline that also includes malignancy classification and metastatic risk prediction. We adopt DeepLabV3+, whose Atrous Spatial Pyramid Pooling (ASPP) captures multi-scale context, with an EfficientNet-B5 backbone for its accuracy-efficiency trade-off. Additionally, we test three channel-attention modules: Squeeze-and-Excitation, CBAM and Grouped Channel Attention (GCA), paired with two loss formulations (Dice-BCE and Boundary).\ Since binary masks do not indicate prediction confidence, we wrap the trained model with a split conformal prediction layer that produces pixel-level uncertainty maps with statistical coverage guarantees.\ We evaluate all configurations on the TN3K public benchmark using a fixed split and directly compare them with state-of-the-art methods. The plain backbone already improves upon previously reported results, with CBAM + Dice-BCE emerging as the best overall configuration.

Figures & tables

Attention Mechanism	Loss Function	Dice $\uparrow$				MCC $\uparrow$
		Dice g	mDice	mDice 50	mDice 10	MCC g	mMCC
None	Dice-BCE	88.76	83.43	89.70	63.83	0.872	0.829
None	Boundary	89.46	84.72	91.19	65.80	0.880	0.843
GCA	Dice-BCE	89.50	84.66	90.54	64.32	0.881	0.842
GCA	Boundary	89.43	84.48	90.70	64.87	0.880	0.840
SE	Dice-BCE	89.18	84.42	90.42	64.51	0.877	0.839

Table 1: Pixel-level segmentation metrics on the TN3K test set (

614

images).

\text{Dice}_{g}

is the pixel-level Dice; mDice ,

\text{mDice}_{50}

\text{mDice}_{10}

are the per-image mean, median and worst

10\%

\text{MCC}_{g}

is the global Matthews Correlation Coefficient and mMCC its per-image mean; both are reported in the

[-1,+1]

scale. Best per column in bold .

Attention Mechanism	Loss Function	DR by size $\uparrow$			DR by count $\uparrow$
		DR s	DR m	DR l	DR 1	DR M
None	Dice-BCE	0.861	0.919	0.970	0.961	0.726
None	Boundary	0.891	0.946	0.980	0.971	0.801
GCA	Dice-BCE	0.883	0.900	0.985	0.958	0.774
GCA	Boundary	0.883	0.923	0.990	0.971	0.767
SE	Dice-BCE	0.898	0.928	0.980	0.963	0.815

Table 2: Detection rate (DR) on the TN3K test set (

614

images). A nodule is detected if at least

50\%

of its area is covered by the prediction.

\text{DR}_{s/m/l}

are the detection rates for small, medium and large nodules (

q_{33}=2255

px,

q_{67}=8062

px);

\text{DR}_{1}

\text{DR}_{M}

are the detection rates for single and multi-nodule images. Best per column in bold .

Figure 1: Qualitative segmentations on two TN3K test cases (one easy, one hard) for the four attention configurations, all trained with Dice-BCE. From left to right: original image, ground truth, prediction.

Figure 2: Conformal prediction output on a TN3K test image. Left: ground-truth nodule (red overlay). Right: conformal prediction inner set (blue) and uncertainty band (green) at coverage level $1-\alpha$ .

Explore similar work

ThyCLIPNet: A BiomedCLIP-Guided Lightweight Attention-Enhanced DeepLabV3+ Framework for Robust Thyroid Nodule Segmentation
Oct 3, 2026Tasnim Jahan, Md Easin Arafat, Swakkhar ShatabdaUltrasound Image Segmentation Medical Image Segmentation
Department of Computer Science and Engineering, United International University, United City, Madani Ave, Dhaka 1212, Bangladesh · Faculty of Informatics, Institute of Industry-Academia Innovation, Department of Data Science and Engineering, Eötvös Lor´and University, P´azm´any P´eter S´et´any 1/C, Budapest 1117, Hungary · Department of Computer Science and Engineering, BRAC University, Pragati Sarani, Merul Badda, Dhaka 1212, Bangladesh
Adversarial Robustness of Deep Learning-Based Thyroid Nodule Segmentation in Ultrasound
Feb 25, 2026Nicholas Dietrich, David McShannonAdversarial Robustness Black-Box Adversarial Attacks
Temerty Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada · Faculty of Engineering, McMaster University, Hamilton, Ontario, Canada
MedSaab-US: A Backpropagation-Free Multi-Scale Wavelet-Saab Framework for Thyroid Nodule Segmentation in Ultrasound Images
Jul 2, 2026Mohammad Amanour RahmanUltrasound Image Segmentation Medical Image Segmentation
Department of Computer Science and Engineering Ahsanullah University of Science and Technology (AUST) Dhaka, Bangladesh

Oct 3, 2026J/K move · Enter open · S save

ThyCLIPNet: A BiomedCLIP-Guided Lightweight Attention-Enhanced DeepLabV3+ Framework for Robust Thyroid Nodule Segmentation

Tasnim Jahan, Md Easin Arafat, Swakkhar Shatabda

Department of Computer Science and Engineering, United International University, United City, Madani Ave, Dhaka 1212, Bangladesh · Faculty of Informatics, Institute of Industry-Academia Innovation, Department of Data Science and Engineering, Eötvös Lor´and University, P´azm´any P´eter S´et´any 1/C, Budapest 1117, Hungary · Department of Computer Science and Engineering, BRAC University, Pragati Sarani, Merul Badda, Dhaka 1212, Bangladesh

Ultrasound Image Segmentation Medical Image Segmentation

Open paper AskPDF arXiv

Accurate thyroid ultrasound segmentation is often challenged by low contrast, speckle noise, and unclear boundaries. Although recent methods have improved segmentation accuracy, many rely on resource-intensive architectures or lack explicit integration of multiscale features with global biomedical visual guidance. In this paper, we introduce ThyCLIPNet, a lightweight semantic-guided hybrid encoder-decoder framework that integrates BiomedCLIP-derived biomedical semantic guidance into a lightweight multi-scale CNN segmentation pipeline. The encoder integrates MobileNetV2 with efficient channel attention, while atrous spatial pyramid pooling and a custom convolutional block attention module enrich bottleneck features. The decoder combines hierarchical skip connections and lightweight attention refinement with a BiomedCLIP-guided gated fusion pathway that projects vision-only global biomedical embeddings into decoder feature space and selectively integrates them through semantic-local fusion and spatial gating. To the best of our knowledge, ThyCLIPNet is among the first lightweight thyroid ultrasound segmentation frameworks to use BiomedCLIP's vision encoder alone for image-only global semantic guidance without text prompting. Experiments on TG3K, TN3K, DDTI, and PKTN achieve dice similarity coefficients of 96.22%, 87.58%, 84.73%, and 80.70%; intersection over union scores of 92.72%, 77.91%, 73.51%, and 67.64%; and 95th-percentile hausdorff distances of 3.75, 16.38, 18.23, and 10.86, respectively. ThyCLIPNet uses 8.55M parameters and 22.99G FLOPs. Overall, the results support integrating global biomedical semantic guidance with lightweight multi-scale CNN representations for robust and computationally efficient thyroid ultrasound segmentation. Source code: https://github.com/Tasnim-Jahan/ThyCLIPNet. [Abstract shortened for arXiv. See PDF for full abstract.]

Abstract

Explore similar work

ThyCLIPNet: A BiomedCLIP-Guided Lightweight Attention-Enhanced DeepLabV3+ Framework for Robust Thyroid Nodule Segmentation

Adversarial Robustness of Deep Learning-Based Thyroid Nodule Segmentation in Ultrasound

MedSaab-US: A Backpropagation-Free Multi-Scale Wavelet-Saab Framework for Thyroid Nodule Segmentation in Ultrasound Images