Deep learning model for the diagnosis of breast cancers smaller than 1 cm with ultrasonography - integration of ultrasonography and clinical factors

Sep 24, 2026

Uploaded by Ivy

Aim of this study is to assess the diagnostic performance of a deep learning algorithm for breast masses smaller than 1 cm on ultrasonography, and to evaluate a hybrid approach that integrates deep learning predictions from ultrasound images with patient clinical factors. The dataset includes 1,041 ultrasonography images from 1,041 patients, split into training, validation, and test sets. A mask region-based CNN with a pre-trained ResNet101 was used, and clinical factors were modeled using an MLP. Results show AUROC of 0.85 for the image-based model, while the combined model using clinical factors outperforms the standalone approach.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Background

Methods

Results

Conclusions

Introduction

Deep learning model for the diagnosis of breast cancers smaller than 1 cm with ultrasonography: integration of ultrasonography and clinical factors Jae Hwan Bong1\#, Tae Hee Kim2\#, Seongkyun Jeong1 1Department of Human Intelligence Robot Engineering, Sangmyung University, Cheonan, South Korea; 2Department of Radiology, Ajou University School of Medicine, Suwon, South Korea Contributions: (I) Conception and design: All authors; (II) Administrative support: S Jeong, TH Kim; (III) Provision of study materials or patients: JHBong, TH Kim; (IV) Collection and assembly of data: TH Kim, JH Bong; (V) Data analysis and interpretation: All authors; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors. \#These authors contributed equally to this work. Correspondence to: Seongkyun Jeong, PhD. Department of Human Intelligence Robot Engineering, Sangmyung University, 31 Sangmyungdae-gil, Dongnam-gu, Cheonan 31066, South Korea. Email: [skjeong@smu.ac.kr](skjeong@smu.ac.kr). Background: The aim of this study was to evaluate the diagnostic performance of a deep learning (DL) algorithm for breast masses smaller than 1 cm on ultrasonography (US) . We also evaluated a hybrid model that combines the predictions of the DL algorithm from US images and a patient’s clinical factors including age, family history of breast cancer, BRCA mutation, and mammographic breast density. Methods: A total of 1,041 US images (including 633 benign and 408 malignant masses) were obtained from 1,041 patients who underwent US between January 2014 and June 2021. All US images were randomly divided into training (513 benign and 288 malignant lesions), validation (60 benign and 60 malignant lesions), and test (60 benign and 60 malignant lesions) data sets. A mask region-based convolutional neural network (R-CNN) was used to generate a feature map of the input image with a CNN and a pre-trained ResNet101 structure. For the clinical model, the multilayer perceptron (MLP) structure was used to calculate the likelihood that the tumor was benign or malignant from the clinical risk factors. We compared the diagnostic performance of an image-based DL algorithm, a combined model with regression, and a combined model with the decision tree method. Results: Using the US images, the area under the receiver operating characteristics curve (AUROC) of the DL algorithm was 0.85 [95% confidence interval (CI), 0.78–0.92] . With the combined model using a regression model, the sensitivity was 78.3%(95% CI, 67.9–88.8%) and the specificity was 85%(95% CI, 76–94%). The sensitivity of the combined model using a regression model was significantly higher than that of the imaging model (P=0.003) . The specificity values of the two models were not significantly different (P=0.083). The sensitivity and specificity of the combined model using a decision tree model were 75%(95% CI, 62.1–85.3%) and 91.7%(95% CI, 81.6–97.2%), respectively. The sensitivity of the combined model using the decision tree model was higher than that of the image model but the difference was not statistically significant (P=0.081). The specificity values of the two models were not significantly different (P=0.748) . Conclusions: The DL model could feasibly be used to predict breast cancers smaller than 1 cm. The combined model using clinical factors outperformed the standalone US-based DL model. Keywords: Breast cancer; deep learning (DL); ultrasonography (US) © Quantitative Imaging in Medicine and Surgery. All rights reserved. Quant Imaging Med Surg 2023;13(4):2486-2495 | [https://dx.doi.org/10.21037/qims-22-880](https://dx.doi.org/10.21037/qims-22-880) Quantitative Imaging in Medicine and Surgery, Vol 13, No 4 April 2023 2487 Submitted Aug 25, 2022. Accepted for publication Feb 11, 2023. Published online Mar 09, 2023. doi: 10.21037/qims-22-880 View this article at: [https://dx.doi.org/10.21037/qims-22-880](https://dx.doi.org/10.21037/qims-22-880) Intr

FAQs

What was the primary goal of this study?

To evaluate how well a deep learning algorithm can diagnose breast masses smaller than 1 cm using ultrasonography, and to test a hybrid model that adds clinical risk factors to image-based predictions.

How was the dataset organized and how many images were used?

A total of 1,041 ultrasonography images were collected from 1,041 patients, and then randomly divided into training, validation, and test sets with separate benign and malignant counts.

What modeling approaches were compared in diagnostic performance?

An image-based deep learning model was compared with combined models that integrate image predictions and clinical factors using regression or a decision tree method.

Sep 24, 2026

Uploaded by Ivy