Deep Learning Based on ACR TI-RADS Can Improve the Differential Diagnosis of Thyroid Nodules

Wu, Ge-Ge; Lv, Wenzhi; Yin, Rui; Xu, Jianwei; Yan, Yichao; Chen, Ruixue; Wang, Jiayu; Zhang, Bo; Cui, Xin-Wu; Dietrich, Christoph F.

doi:10.3389/fonc.2021.575166

Cited by 34 publications

(27 citation statements)

References 36 publications

(47 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In cases of indeterminate lesions by conventional US and in cases of diagnostic inconsistencies between CEUS and SWE, combined scores and further fine-needle aspiration biopsy is recommended to improve diagnostic accuracy. Furthermore, many recent studies suggest that artificial intelligence algorithms increase the accuracy of diagnosing benign versus malignant thyroid nodules, especially in the TI-RADS 4 and 5 categories, and help reduce the rate of unnecessary FNAB from 62% to 35% [111][112][113]. In 2020, Xu et al published a meta-analysis which included 19 papers with 4781 thyroid nodules, analyzing the performance of Computer Aided Diagnosis (CAD) systems performance in differentiation of malignant thyroid nodules: the deep learning-based system showed a sensitivity of 87% and a specificity of 85%.…”

Section: Discussionmentioning

confidence: 99%

Performance of Contrast-Enhanced Ultrasound in Thyroid Nodules: Review of Current State and Future Perspectives

Radziņa

Ratniece

Putriņš

et al. 2021

Cancers

View full text Add to dashboard Cite

Ultrasound has been established as a baseline imaging technique for thyroid nodules. The main advantage of adding CEUS is the ability to assess the sequence and intensity of vascular perfusion and hemodynamics in the thyroid nodule, thus providing real-time characterization of nodule features, considered a valuable new approach in the determination of benign vs. malignant nodules. Original studies, reviews and six meta-analyses were included in this article. A total of 624 studies were retrieved, and 107 were included in the study. As recognized for thyroid nodule malignancy risk stratification by US, for acceptable accuracy in malignancy a combination of several CEUS parameters should be applied: hypo-enhancement, heterogeneous, peripheral irregular enhancement in combination with internal enhancement patterns, and slow wash-in and wash-out curve lower than in normal thyroid tissue. In contrast, homogeneous, intense enhancement with smooth rim enhancement and “fast-in and slow-out” are indicative of the benignity of the thyroid nodule. Even though overlapping features require standardization, with further research, CEUS may achieve reliable performance in detecting or excluding thyroid cancer. It can also play an operative role in guiding ablation procedures of benign and malignant thyroid nodules and metastatic lymph nodes, and providing accurate follow-up imaging to assess treatment efficacy.

show abstract

Section: Discussionmentioning

confidence: 99%

Performance of Contrast-Enhanced Ultrasound in Thyroid Nodules: Review of Current State and Future Perspectives

Radziņa

Ratniece

Putriņš

et al. 2021

Cancers

View full text Add to dashboard Cite

show abstract

“…In addition, the ML-based investigations reported in [ 10 , 22 ] have introduced a semi-automatic method that is characterized by an initial automatic selection of a box region and subsequently by a manual contouring performed by expert clinicians. Conversely, the studies that applied DL algorithms to thyroid imaging considered a manual selected box around the region under investigation [ 9 , 11 , 52 , 54 ]. Furthermore, it is worth pointing out that radiomics studies are based on a manual contouring along the borders of the thyroid tumor [ 60 , 61 , 62 ] or slightly within the borders of the tumor to avoid artifacts [ 64 ].…”

Section: Discussionmentioning

confidence: 99%

“…Mainly, AI algorithms have been implemented for the classification of thyroid nodules, i.e., differentiating among benign or malignant state [ 9 , 10 , 21 , 22 , 33 , 41 , 51 , 52 , 53 , 54 , 55 , 56 ]. The outcomes of these studies are compared with the diagnosis of radiologists with different levels of experience.…”

Section: Ai and Radiomics In Thyroid Diseasesmentioning

confidence: 99%

Artificial Intelligence in Thyroid Field—A Comprehensive Review

Bini

Pica

Azzimonti

et al. 2021

Cancers

View full text Add to dashboard Cite

Artificial intelligence (AI) uses mathematical algorithms to perform tasks that require human cognitive abilities. AI-based methodologies, e.g., machine learning and deep learning, as well as the recently developed research field of radiomics have noticeable potential to transform medical diagnostics. AI-based techniques applied to medical imaging allow to detect biological abnormalities, to diagnostic neoplasms or to predict the response to treatment. Nonetheless, the diagnostic accuracy of these methods is still a matter of debate. In this article, we first illustrate the key concepts and workflow characteristics of machine learning, deep learning and radiomics. We outline considerations regarding data input requirements, differences among these methodologies and their limitations. Subsequently, a concise overview is presented regarding the application of AI methods to the evaluation of thyroid images. We developed a critical discussion concerning limits and open challenges that should be addressed before the translation of AI techniques to the broad clinical use. Clarification of the pitfalls of AI-based techniques results crucial in order to ensure the optimal application for each patient.

show abstract

“…We collected 3906 fully anonymized images of thyroid nodule lesions from 3906 patients of different ages (mean age 48 years) patients from the Cancer Hospital of Tianjin Medical University, and this part of the data is called the internal dataset. In our study, the ultrasound images used for deep learning should satisfy the following criteria: (1) no color Doppler flow images, (2) images with only one thyroid nodule, (3) images with a clear thyroid nodule, and (4) available surgical pathological results 15 . All ultrasound images have been carefully reviewed by expert physicians and any images that do not meet the criteria are excluded.…”

Section: Methodsmentioning

confidence: 99%

Multitask network for thyroid nodule diagnosis based on TI‐RADS

et al. 2022

View full text Add to dashboard Cite

Purpose Assessment of thyroid nodules is usually relied on the experience of the radiologist and is time‐consuming. Classification model of thyroid nodules cannot only reduce the burden on physicians but also provide objective recommendations. However, most classification models based on deep learning simply give a prediction result of the benignity or malignancy of nodules; thus, physicians have no way of knowing how the deep learning gets the prediction result due to the black‐box nature of neural networks. In this work, we integrate the explainability directly into the outputs generated by the model through combining thyroid imaging reporting and data system (TI‐RADS). The inference process of the proposed method is consistent with doctor's clinical diagnosis process; therefore, doctors can better explain the diagnosis results of the model to the patient. Methods A multitask network based on TI‐RADS (MTN‐TI‐RADS) for the classification of thyroid nodules is proposed. In this network, a set of TI‐RADS classifications of nodules is first obtained by multitask learning, then the TI‐RADS points and the corresponding risk levels are calculated, and finally, nodules are classified as benign and malignant. The classification process through the network is consistent with the diagnostic process of physician; thus, the results of classification can be easily understood by physicians. In addition, the attention modules are introduced to the spatial and channel domains to let the network focus more on critical features. Results To verify the classification performance of our method, we compared the results obtained through our method with the results of the radiologist's evaluation. For the 781 test nodules in the internal dataset and the 886 test nodules in the external dataset, the sensitivity and specificity of MTN‐TI‐RADS were 0.988, 0.912 in internal dataset, 0.949, 0.930 in external dataset, versus the senior radiologist of 0.925 (p<0.001$p<0.001$), 0.816 (p=0.005$p=0.005$), and 0.910 (p=0.009$p=0.009$), 0.836 (p<0.001$p<0.001$), respectively. And the area under the receiver operating characteristic curve of MTN‐TI‐RADS was 0.981 in internal dataset, 0.973 in external dataset, versus the senior radiologist of 0.905, 0.923. For the internal dataset, we also computed the accuracy of the risk level (TR1 to TR5) and the mean absolute error (MAE). The accuracy of the risk level of the proposed method is 78%, and the MAE is 1.30. The MAE of the total points (0–14 points) is 1.30. Conclusions An effective and result‐interpretable end‐to‐end thyroid nodule classification network (MTN‐TI‐RADS) is proposed. MTN‐TI‐RADS has superior ability to classify malignant and benign thyroid nodules compared to senior radiologists. Based on MTN‐TI‐RADS, a classification model with strong interpretation and a high degree of physician trust is constructed. The proposed classification network is consistent with the diagnosis process of physicians, thus is more reliable and interpretable, and has great potential for clinical ...

show abstract

Deep Learning Based on ACR TI-RADS Can Improve the Differential Diagnosis of Thyroid Nodules

Cited by 34 publications

References 36 publications

Performance of Contrast-Enhanced Ultrasound in Thyroid Nodules: Review of Current State and Future Perspectives

Performance of Contrast-Enhanced Ultrasound in Thyroid Nodules: Review of Current State and Future Perspectives

Artificial Intelligence in Thyroid Field—A Comprehensive Review

Multitask network for thyroid nodule diagnosis based on TI‐RADS

Contact Info

Product

Resources

About