The diagnostic performance of ultrasonic S-Detect technology in the diagnosis of BI-RADS-4 breast nodules smaller and larger than 20 mm

doi:10.21203/rs.3.rs-3608096/v1

Download PDF

Research Article

The diagnostic performance of ultrasonic S-Detect technology in the diagnosis of BI-RADS-4 breast nodules smaller and larger than 20 mm

https://doi.org/10.21203/rs.3.rs-3608096/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background: We aimed to explore the value of ultrasonic S-Detect in the differential diagnosis of Breast Imaging-Reporting and Data System (BI-RADS) 4 breast nodules smaller and larger than 20 mm.

Methods: Between November 2020 and November 2022, a total of 382 breast nodules in 312 patients were classified as BI-RADS-4 by conventional ultrasound. Using pathology results as the gold standard, we applied receiver operator characteristics (ROC), sensitivity (SE), specificity (SP), accuracy (ACC), positive predictive value (PPV), and negative predictive value (NPV) to analyze the diagnostic value of BI-RADS, S-Detect, and the two techniques in combination (Co-Detect) in the diagnosis of BI-RADS 4 breast nodules smaller and larger than 20 mm.

Results: There were 382 BI-RADS-4 nodules, of which 151 were pathologically confirmed as malignant, and 231 as benign. In lesions ≤ 20 mm, the SE, SP, ACC, PPV, NPV, area under the curve (AUC) of the BI-RADS group were 77.27%, 89.73%, 85.71%, 78.16%, 89.25%, 0.835, respectively. SE, SP, ACC, PPV, NPV, AUC of the S-Detect group were 95.05%, 78.92%, 83.15%, 67.50%, 95.42%, 0.855, respectively. SE, SP, ACC, PPV, NPV, AUC of the Co-Detect group were 89.77%, 93.51%, 92.31%, 86.81%, 95.05%, 0.916, respectively. The difference of SE, ACC, NPV, AUC between the BI-RADS group and the Co-Detect group was statistically significant (P < 0.05). In lesions > 20 mm, SE, SP, ACC, PPV, NPV, AUC of the BI-RADS group were 88.99%, 89.13%, 88.99%, 91.80%, 85.42%, 0.890, respectively. SE, SP, ACC, PPV, NPV, AUC of the S-Detect group were 98.41%, 69.57%, 86.24%, 81.58%, 96.97%, 0.840, respectively. SE, SP, ACC, PPV, NPV, AUC of the Co-Detect group were 98.41%, 91.30%, 95.41%, 93.94%, 97.67%, 0.949 respectively. The difference in AUC between the BI-RADS group and the Co-Detect group was statistically significant (P < 0.05).

Conclusions: S-Detect combined with BI-RADS is highly effective in the differential diagnosis of BI-RADS 4 breast nodules smaller and larger than 20 mm. However, being especially effective for lesions ≤ 20 mm, S-Detect combined with BI-RADS allows to reduce a number of unnecessary biopsies.

Ultrasound

Artificial intelligence

S-Detect

Breast mass

Breast imaging reporting and data system

Ranking as the second leading cause of cancer-related death among women, breast cancer has become the biggest threat to their health ¹. The American College of Radiology has published the fifth edition of Breast Imaging-Reporting and Data System (BI-RADS) which is a set of guidelines on breast ultrasound imaging standards. According to one of them, patients with BI-RADS 4 breast nodules, with a malignancy rate of 3–94%, need to have a biopsy ^2,3. The biopsy is sometimes unnecessary. Although widely applied in breast cancer diagnosis, BI-RADS is influenced by subjective factors and differences between observers, and, as such, does not always accurately assess whether the nodule is benign or malignant.

“S-Detect”, a computer-aided diagnosis (CAD) technology is a software program based on a convolutional neural network developed by Samsung. Based on the breast imaging reporting data recommended by BI-RADS, S-Detect automatically detects and analyzes the internal structure, boundary, location, shape, and other features of breast nodules using a deep learning model. The analysis enables differentiation between benign and malignant lesions, without a bias resulting from a subjective assessment of the physician ⁴. In the literature to date, there have not been many reports on the influence of nodule size on the diagnostic performance and guiding application in early breast cancer using CAD technology ^5,6,7.

Thus, we used S-Detect to further diagnose breast nodules, which were classified as BI-RADS 4 by conventional ultrasound, and compared the diagnostic performance of S-Detect combined with BI-RADS (Co-Detect) for BI-RADS 4 breast nodules, both smaller and larger than 20 mm. To improve the diagnostic accuracy of benign and malignant BI-RADS 4 breast nodules, we also investigated the influence of the size of nodules on the diagnostic performance of Co-Detect.

Patients Selection

Inclusion criteria: (1) one or more nodules in the breast; (2) a coarse needle biopsy or surgical removal as the initial treatment, (3) a definitive pathological diagnosis by biopsy or conventional surgical excision. Exclusion criteria: (1) anamnesis and prior breast tissue biopsy or surgery; (2) cystic lesions or calcifications; (3) the nodule whose boundary was not identified by S-Detect because the nodules were too small or too large.

Finally, from November 2020 to November 2022, 382 nodules were detected in 302 women (age range: 12–90 years) from the Yichang Central People’s Hospital. Institutional review board approval from the Yichang Central People’s Hospital was obtained for this study (PJ-KY2019-11).

Instruments and strategies

A Samsung RS80A diagnostic ultrasound machine with an L3-12A probe (5–13 MHz) was used. The patient was placed in the supine position to expose the mammary glands area as much as possible, and multi-slice images from each quadrant were acquired. The images were adjusted to obtain the maximal diameter of the nodule, and the relationship between the nodule and the adjacent tissue. The information on the size, shape, location, boundary, internal echo, calcification, blood flow, and homolateral axillary lymph node was collected.

Nodules were described and classified based on the breast imaging reporting data recommended by BI-RADS. In addition, the images were adjusted to acquire longitudinal and transverse sections along the largest diameter and to show the relationship between the nodule and the adjacent tissue. Then the S-Detect mammary gland mode was selected to automatically outline the boundaries of the nodule, and the operator could readjust the frame, if necessary. The software provided a categorization into: possibly malignant or possibly benign.

Diagnostic Criteria

BI-RADS

According to the 2013 version of the BI-RADS classification, the shape, margin and orientation of the nodule are important for establishing an accurate diagnosis, while the internal echo pattern, posterior acoustic features and blood flow information are considered less significant. Calcification classification is changed from macrocalcifications and microcalcifications to internal, peripheral, and intraductal calcifications, wherein internal and intraductal calcifications indicate malignancy.

BI-RADS 3 is considered probably benign (≤ 2% likelihood of malignancy). Images with malignant signs are categorized as BI-RADS 4, which is assessed as a possible malignant tumor (3% − 94% likelihood of malignancy) and is further subdivided into BI-RADS 4a, 4b and 4c. Malignancy is suggested by signs that include an irregular shape, non-circumscribed margin, antiparallel orientation, posterior shadowing, microcalcifications, and abnormalities in the surrounding tissue. Images with one, two, and three of these features are categorized as 4a, 4b, and 4c, respectively; BI-RADS 5 indicates a high malignancy rate (> 95% likelihood of malignancy). BI-RADS 6 is confirmed malignant by biopsy ^8,9. Accordingly, BI-RADS ≤ 4a is assessed as probably benign, and BI-RADS ≥ 4b as probably malignant.

S-Detect

The longitudinal and transverse sections along the largest diameter were tested using S-Detect, which automatically generated diagnostic reports, defining a given lesion as probably benign or probably malignant. If the prompt results of any section differed, the nodule was considered probably malignant. If the prompt results of two sections indicated possible benign tumors, the lesion was considered probably benign.

Co-Detect

If the classification results of the longitudinal and transverse sections tested by S-Detect were both possibly malignant, the original BI-RADS diagnostic results were upgraded by one level. For example, BI-RADS 4a was upgraded to BI-RADS 4b. When the results were both possibly benign, the original BI-RADS diagnostic results were reduced by one level. In contrast, if the classification results of the longitudinal and transverse sections tested by S-Detect were different, the original BI-RADS diagnostic results were invariant, BI-RADS ≥ 4b was assessed as probably malignant, and BI-RADS ≥ 4a as probably benign.

All nodules smaller and larger than 20 mm were assessed by two ultrasonographers who received professional training and had more than five years of experience. In case of conflicting findings, senior ultrasound doctors with more than ten years of experience jointly reviewed the images, discussed the results, and achieved a consensus.

Statistical Analysis

Statistical evaluation was performed using SPSS version 23.0 and MedCalc 19.6.4. Quantitative data were expressed as mean ± standard deviation/error, and enumeration data were compared by the χ2 test. Using pathological results as the gold standard, we analyzed the difference in diagnostic sensitivity (SE), specificity (SP), accuracy (ACC), positive predictive value (PPV), and negative predictive value (NPV) between the groups. The receiver operator characteristics (ROC) curves were drawn and the under curve area (AUC) of different diagnostic methods was calculated. The AUC was compared by the Z test. P < 0.05 was considered to indicate statistical significance.

General information

Of the 382 nodules, 151 were malignant, and 231 were benign. The age range was between 12 and 90. The patients with malignant nodules had an average age of 52.76 ± 11.80, and the patients with benign nodules had an average age of 43.31 ± 11.07 years, the difference was obvious (P < 0.001). The largest diameter of the nodules in total was 2.60–48.80 mm. And the largest diameter of malignant nodules was 19.00 ± 8.47 mm, and of benign nodules was 14.57 ± 7.61 mm, the difference was significant (P < 0.001) (Table 1).

Table 1

Patient demographics and lesion characteristics.
Variables	Malignant Mass (n = 151)	Benign Mass (n = 231)	p
Variables	Number (%)	Number (%)	p
Patient age (years), mean ± SD	52.76 ± 11.80	43.31 ± 11.07	< 0.001
Lesion size (mm), mean ± SD	19.00 ± 8.47	14.57 ± 7.61	< 0.001
Histopathologic diagnosis
Malignant lesions (n = 151)
Invasive ductal carcinoma	124 (82.1)
Intraductal carcinoma in situ	11 (7.3)
Mucinous adenocarcinoma	5 (3.3)
Invasive lobular carcinoma	4 (2.6)
Papillocarcinoma	2 (1.3)
Adenoid cystic carcinoma	2 (1.3)
Intraductal papillary carcinoma	1 (0.7)
encapsulated papillary carcinoma	1 (0.7)
Metaplastic carcinoma	1 (0.7)
Benign lesions (n = 231)
Fibroadenosis with fibroadenoma		102 (44.2)
Fibroadenosis	v	70 (30.3)
Fibroadenoma		24 (10.4)
Intraductal papilloma		17 (7.4)
Granulomatous lobulitis		12 (5.2)
Sclerosing adenosis		4 (1.7)
Fibrocystic mammary disease		1 (0.4)
Benign phyllodes		1 (0.4)
ACR BI-RADS category
4a	27 (17.9)	207 (89.6)
4b	65 (43.0)	23 (10.0)
4c	59 (39.1)	1 (0.4)

Diagnostic results of BI-RADS and Co-Detect

273 lesions were ≤ 20 mm, in the BI-RADS group, 186 lesions assigned to BI-RADS 4a (188) were benign; 87 lesions assigned to BI-RADS 4b (56), and BI-RADS 4c (31) were malignant. In the Co-Detect group, 182 lesions assigned to BI-RADS 3 (138) and BI-RADS 4a (44) were benign; 91 lesions assigned to BI-RADS 4b (27), BI-RADS 4c (37), and BI-RADS 5 (27) were malignant. 252 lesions were diagnosed correctly (BI-RADS 3 (132), BI-RADS 4a (41), BI-RADS 4b (19), BI-RADS 4c (33), BI-RADS 5 (27)). The ACC was improved from85.71% (234/273) to 92.31% (252/273).

109 lesions were > 20 mm, in the BI-RADS group, 48 lesions assigned to BI-RADS 4a (48) were benign; 61 lesions assigned to BI-RADS 4b (32) and BI-RADS 4c (29) were malignant. In the Co-Detect group, 43 lesions assigned to BI-RADS 3 (29) and BI-RADS 4a (14) were benign; 66 lesions assigned to BI-RADS 4b (12), BI-RADS 4c (28), and BI-RADS 5 (26) were malignant. 104 lesions were diagnosed correctly (BI-RADS 3 (29), BI-RADS 4a (13), BI-RADS 4b (11), BI-RADS 4c (25), BI-RADS 5 (26)). The ACC was improved from 88.99% (97/109) to 95.41% (104/109) (Tables 2, 3).

Table 2

Comparison of ultrasonic diagnosis of BI-RADS before and after adjustment by S-Detect technology.
Pathological type	BI-RADS scores (lesions ≤ 20 mm)			BI-RADS scores (lesions > 20 mm)			BI-RADS Post-adjustment (lesions ≤ 20 mm)					BI-RADS Post-adjustment (lesions > 20 mm)
	4a	4b	4c	4a	4b	4c	3	4a	4b	4c	5	3	4a	4b	4c	5
Malignant	20	37	31	7	28	28	6	3	19	33	27	0	1	11	25	26
Benign	166	19	0	41	4	1	132	41	8	4	0	29	13	1	3	0
合计	188	56	31	48	32	29	138	44	27	37	27	29	14	12	28	26

Table 3

Comparison of ultrasonic diagnosis among the BI-RADS, S-Detect, and Co-Detect groups with pathology results.
Method	lesions ≤ 20 mm		lesions > 20 mm
Method	Malignant	Benign	Malignant	Benign
BI-RADS
Malignant	68	19	56	5
Benign	20	166	7	41
S-Detect
Malignant	81	39	62	14
Benign	7	146	1	32
Co-Detect
Malignant	79	12	62	4
Benign	9	173	1	42

Diagnostic performance of BI-RADS, S-Detect and Co-Detect

For all breast nodules smaller and larger than 20mm, the Co-Detect group had improved SE, SP, ACC, PPV, and NPV. The difference in SE, ACC, and NPV between the BI-RADS group and the Co-Detect group was statistically significant (P < 0.05). For breast nodules ≤ 20 mm, the difference in SP, ACC, and PPV between the S-Detect group and the Co-Detect group was statistically significant (P < 0.001). For breast nodules > 20 mm, the difference of SE, SP, ACC, PPV, and NPV between the BI-RADS group and the Co-Detect group was not statistically significant (P > 0.05) (Table 4).

Table 4

Comparison of the performance of the BI-RADS, S-Detect, and Co-Detect groups in diagnosing BI-RADS 4 breast nodules smaller and larger than 20 mm.
	Lesions ≤ 20 mm			Lesions > 20 mm
	BI-RADS	S-Detect	Co-Detect	BI-RADS	S-Detect	Co-Detect
SE	77.27	95.05^*	89.77^*	88.89	98.41	98.41
SP	89.73	78.92	93.51^#	89.13	69.57	91.30^#
ACC	85.71	83.15	92.31^*	88.99	86.24	95.41
PPV	78.16	67.50	86.81^*	91.80	81.58	93.94^#
NPV	89.25	95.42^*	95.05^*	85.42	96.97	97.67
AUC	0.835	0.855^*	0.916^*#	0.890	0.840	0.949^*#

Comparison with the BI-RADS group, ^*P༜0.05; Comparison with the S-Detect group, ^#P༜0.05

The results of ROCs are shown in Figure. 1. For breast nodules ≤ 20 mm, the AUC values of the BI-RADS, S-Detect, and Co-Detect groups were 0.835, 0.855, and 0.916, respectively. The AUC value of the Co-Detect group was higher than that of the BI-RADS and S-Detect groups (BI-RADS vs. Co-Detect, Z = 3.632, p < 0.001; S-Detect vs. Co-Detect, Z = 4.035, P < 0.001). For breast nodules > 20 mm, the AUC values of the BI-RADS, S-Detect, and Co-Detect groups were 0.890, 0.840, and 0.949, respectively. The AUC value of the Co-Detect group was higher than that of the BI-RADS and S-Detect groups (BI-RADS vs. Co-Detect, Z = 2.024, p = 0.043; S-Detect vs. Co-Detect, Z = 3.536, P < 0.001)

Ultrasonography is an important technique to screen for breast lesions. BI-RADS has been developed to describe the characteristics of these lesions in order to categorize them according to the likelihood of malignancy. According to the 2013 version of the BI-RADS classification, BI-RADS 4 is further subdivided into BI-RADS 4a, 4b, and 4c. BI-RADS 4a is assessed as a possible benign tumor (3% − 10% likelihood of malignancy), in which case an aspiration biopsy or a short-interval follow-up is recommended. BI-RADS 4b is assessed as a possible malignant tumor (11% − 50% likelihood of malignancy), in which case a biopsy should be performed. BI-RADS 4c indicates a high malignancy rate (51% − 94% likelihood of malignancy), for which a biopsy or surgery is indicated ^8,9. However, all nodules classified into the BI-RADS 4 category, without further division into 4a, 4b, and 4c subgroups, are predicted to have a 3–94% malignancy rate. Not only does such a wide range pose the risk of unnecessary biopsies or surgeries, but also involves great variability in how breast nodules are treated ^10,11. In breast cancer, the size of nodules is an important prognostic index. The largest diameter of early breast cancer is usually ≤ 20 mm, of which the nodules with a diameter ≤ 10 mm are referred to as microcarcinoma. Not only the size of the nodule, however, determines an accurate diagnosis. The performance of BI-RADS in diagnosing early breast cancer may be compromised by atypical image features specific to early breast cancer, overlapping ultrasound images ^12,13. Based on the features of benign and malignant breast nodules, S-Detect processed specific parameters from ultrasound images to accurately identify abnormalities. Then S-Detect automatically identified the lesion as probably benign or probably malignant, by analyzing the size, shape, boundary, retrotumoral echo of the given breast nodule Automation provided by S-Detect eliminates the bias arising from subjective physicians’ assessments ¹⁴ and has shown higher performance in diagnosing breast nodules, as proven by other studies ¹⁵. Nevertheless, studies on S-detect as an additional technique improving the diagnostic performance of BI-RADS 4 breast nodules with different sizes is need to be further studied.

Herein, we found that Co-Detect (BI-RADS/S-Detect in combination) could have higher diagnostic performance and that this performance varied with the size of nodules. Specifically, for breast nodules ≤ 20 mm, SE, ACC, NPV, and AUC were improved in the Co-Detect group, with a statistically significant difference (P < 0.05). For breast nodules ≤ 20 mm, SE, SP, ACC, PPV, and NPV were improved in the Co-Detect group, but the difference was not statistically significant (P < 0.05). We indicated that the diagnostic performance of Co-Detect was associated with the size of nodules, especially with the size ≤ 20 mm. The possible reasons include small lesions in early breast cancer, a weak interstitial response, a loose arrangement of fibrous tissues, fewer new blood vessels, and tumor cells not infiltrating the periphery of the lesion. Thus, it is difficult to differentiate between benign and malignant nodules with conventional ultrasound imaging, especially for physicians with insufficient experience. A 53-year-old female patient included in this study had a 5.4 × 6.5 mm nodule in her right breast (Fig. 2). The ultrasound scan showed a clear boundary, irregular shape, uniform echo, and the longitudinal/transverse ratio > 1. The nodule was classified as BI-RADS 4a by two ultrasonographers. The classification results based on longitudinal and transverse sections obtained with S-Detect were both possibly malignant. Therefore, this nodule was finally diagnosed as a malignant tumor which proved to be an invasive lobular carcinoma, as confirmed by histopathology. Another 39-year-old female patient had an 11.0 x 8.2 mm nodule in her left breast (Fig. 3). The ultrasound scan showed a clear boundary, irregular shape, uniform echo, and antiparallel orientation. The nodule was also classified as BI-RADS 4a by two ultrasonographers. The classification results based on longitudinal and transverse sections obtained with S-Detect were both possibly benign. Thus, this nodule was finally diagnosed as a benign tumor which proved to be an intraductal papilloma by histology. The results obtained with S-Detect became reliable medical recommendations to clinicians on diagnosing benign and malignant nodules from the BI-RADS 4 category, especially breast nodules ≤ 20mm. According to the BI-RADS, S-Detect standardly evaluated and differentiated benign from malignant breast nodules. S-Detect provided a comprehensive practical reference and avoided a bias resulting from a subjective assessment of the physician, thus improving the diagnostic efficiency for breast nodules and playing a more effective role in assisting diagnosis ^16,17,18. For breast nodules > 20 mm, due to the long growth time, some nodules have typical malignant characteristics, including internal liquefaction and calcification. Hence it is easy to differentiate benign breast nodules larger than 20 mm from malignant lesions by conventional ultrasound. Thus, the role of S-Detect in enhancing diagnostic efficacy for breast nodules larger than 20 mm was less significant.

Compared with the studies using S-Detect to diagnose breast nodules, we diagnosed BI-RADS 4 nodules of different sizes. SE, SP, ACC, PPV, NPV, and AUC were used to assess the diagnostic value of S-Detect for BI-RADS 4 nodules of different sizes ^19,20,21. We found that S-Detect could use the difference in nodule size to differentiate benign from malignant BI-RADS 4 breast nodules. The diagnostic coincidence rate of S-Detect was considerably higher than that of conventional ultrasound diagnosis, providing comprehensive guidance in clinical practice. A study has revealed that the use of S-Detect alone or overreliance on S-Detect technology can cause missed diagnoses and misdiagnoses in clinical practice. At present, S-Detect is recommended only as an auxiliary diagnostic tool, rather than as a single or definitive diagnostic method ^22,23,24.

Although S-Detect had a high diagnostic efficacy for BI-RADS 4 breast nodules, there were still certain limitations to the study: (1) This was a single-center study and there was only one type of pathological result examined. Most of the malignant nodules were invasive ductal carcinoma, and most of the benign nodules were fibroadenosis with or without fibroadenoma, with other pathological types being rare; (2) Breast nodules were three-dimensional, while the diagnosis made with S-Detect technology was based mainly on analyzing static images of nodules obtained from the maximum section and vertical section. Therefore, there was a degree of subjectivity in the assessment, although sonographers analyzed the most representative section of the lesion. In subsequent studies, the sample size should be increased for greater validation of the results.

In conclusion, S-Detect combined with BI-RADS could improve the diagnosis of BI-RADS 4 breast nodules, especially nodules ≤ 20 mm, and significantly reduce the rates of missed diagnosis and misdiagnosis caused by subjective assessment of an inexperienced sonographer, thereby reducing the number of unnecessary biopsies. This technique is an auxiliary tool for the objective description and accurate diagnosis of breast nodules and is of significant clinical application value.

BI-RADS

Breast Imaging-Reporting and Data System

CAD

computer-aided diagnosis

sensitivity

specificity

ACC

accuracy

PPV

positive predictive value

NPV

negative predictive value

ROC

receiver operator characteristics

AUC

under curve area.

Acknowledgements

Not applicable.

Authors' contributions

Conception and design: BX, CF; Administrative support: BX, BZ; Provision of study materials or patients: BX, CF, BZ; Collection and assembly of data: BX, CG, YT; Data analysis and interpretation: BX, YT. Manuscript completion: BX, YT. All authors contributed to the article and approved the submitted version.

Funding

The Scientifific Research Program of Education Department of Hubei Province (No. B2022030), Open Fund Project of the Third-grade Pharmacological Laboratory on Traditional Chinese Medicine,State Administration of Traditional Chinese Medicine, China Three Gorges University (2023PTCM09)

Availability of data and materials

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Ethics approval and consent to participate

Institutional review board approval from the Yichang Central People’s Hospital was obtained for this study (PJ-KY2019-11). Each patient signed an informed consent form in our research.

Consent for publication

Not applicable.

Competing interests

The authors have no conflicts of interest to declare.

Author details

¹Department of Ultrasound, Yichang Central People's Hospital, The First College of Clinical Medical Science, China Three Gorges University, Yichang, China. ²Department of Medical Ultrasound, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China. ³Third-grade Pharmacological Laboratory on Traditional Chinese Medicine, State Administration of Traditional Chinese Medicine, China Three Gorges University, Yichang, China

Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209–49.
Dorsi CJ, Sickles EA, Mendelson EB, et al. ACR BI-RADS atlas, breast imaging reporting and data system. 5th ed. Reston: American College of Radiology; 2013. pp. 123–32.
Spak DA, Plaxco JS, Santiago L, et al. BI-RADS® fifth edition: A summary of change. Diagn Interv Imaging. 2017;98(3):179–90.
Chen P, Tong J, Lin T, et al. The added value of S-detect in the diagnostic accuracy of breast masses by senior and junior radiologist groups: a systematic review and meta-analysis. Gland Surg. 2022;11(12):1946–60.
Wang Q, Chen H, Luo G, et al. Performance of novel deep learning network with the incorporation of the automatic segmentation network for diagnosis of breast cancer in automated breast ultrasound. Eur Radiol. 2022;32(10):7163–72.
Al-Antari MA, Al-Masni MA, Kim TS. Deep learning computer-aided diagnosis for breast lesion in digital mammogram. Adv Exp Med Biol. 2020;1213:59–72.
Nicosia L, Addante F, Bozzini AC, et al. Evaluation of computer-aided diagnosis in breast ultrasonography: Improvement in diagnostic performance of inexperienced radiologists. Clin Imaging. 2022;82:150–5.
Liu H, Chen Y, Zhang Y, Wang L, et al. A deep learning model integrating mammography and clinical factors facilitates the malignancy prediction of BI-RADS 4 microcalcifications in breast cancer screening. Eur Radiol. 2021;31(8):5902–12.
Gu Y, Tian JW, Ran HT, et al. The utility of the fifth edition of the bi-rads ultrasound lexicon in category 4 breast lesions: a prospective multicenter study in china. Acad Radiol. 2022;29:26–34.
Spinelli Varella MA, Teixeira da Cruz J, Rauber A, et al. Role of BI-RADS ultrasound subcategories 4a to 4c in predicting breast cancer. Clin Breast Cancer. 2018;18(4):e507–11.
He P, Cui LG, Chen W, Yang RL. Subcategorization of ultrasonographic BI-RADS category 4: assessment of diagnostic accuracy in diagnosing breast lesions and influence of clinical factors on positive predictive value. Ultrasound Med Biol. 2019;45(5):1253–8.
Forsare C, Bak M, Falck AK, et al. Non-linear transformations of age at diagnosis, tumor size, and number of positive lymph nodes in prediction of clinical outcome in breast cance. BMC Cancer. 2018;18(1):1226.
Shao S, Yao M, Li C, et al. Ultrasound features for prediction of long-term outcomes of women with primary breast cancer < 20mm. Front Oncol. 2023;16:1103397.
Zhao C, Xiao M, Jiang Y, et al. Feasibility of computerassisted diagnosis for breast ultrasound: The results of the diagnostic performance of s-detect from a single center in china. Cancer Manag Res. 2019;11:921–30.
Xing B, Chen X, Wang Y, et al. Evaluating breast ultrasound S-detect image analysis for small focal breast lesions. Front Oncol. 2022;13:1030624.
Moon WK, Lee YW, Ke HH, et al. Computer-aided diagnosis of breast ultrasound images using ensemble learning from convolutional neural networks. Comput Methods Programs Biomed. 2020;190:105361.
Morgan MB, Mates JL. Applications of artificial intelligence in breast imaging. Radiol Clin North Am. 2021;59(1):139–48.
Choi JH, Kang BJ, Baek JE, Lee HS, Kim SH. Application of computer-aided diagnosis in breast ultrasound interpretation: improvements in diagnostic performance according to reader experience. Ultrasonography. 2018;37(3):217–25.
Lee J, Kim S, Kang BJ, Kim SH, Park GE. Evaluation of the effect of computer aided diagnosis system on breast ultrasound for inexperienced radiologists in describing and determining breast lesions. Med Ultrason. 2019;21(3):239–45.
Choi JH, Kang BJ, Baek JE, Lee HS, Kim SH. Application of computer-aided diagnosis in breast ultrasound interpretation: improvements in diagnostic performance according to reader experience. Ultrasonography. 2018;37(3):217–25.
Wu JY, Zhao ZZ, Zhang WY, et al. Computer-aided diagnosis of solid breast lesions with ultrasound: factors associated with false-negative and false-positive results. J Ultrasound Med. 2019;38(12):3193–202.
Zhao F, Xiao D, Wen H, et al. Application of computer-aidedultrasonography in assisting radiologists to diagnose early breastcancer. J of Molecular Imaging. 2021;44(2):226–31.
Chambara N, Ying M. The Diagnostic efficiency of ultrasound computer-aided diagnosis in differentiating thyroid nodules: a systematic review and narrative synthesis. Cancers (Basel). 2019;8(11):1759.
Wang XY, Cui LG, Feng J, Chen W. Artificial intelligence for breast ultrasound: An adjunct tool to reduce excessive lesion biopsy. Eur J Radiol. 2021;138:109624.

No competing interests reported.

Download PDF

Reviewers agreed at journal
18 Apr, 2024
Reviewers agreed at journal
13 Feb, 2024
Reviewers invited by journal
11 Dec, 2023
Editor assigned by journal
11 Dec, 2023
Editor invited by journal
01 Dec, 2023
Submission checks completed at journal
01 Dec, 2023
First submitted to journal
13 Nov, 2023

You are reading this latest preprint version

The diagnostic performance of ultrasonic S-Detect technology in the diagnosis of BI-RADS-4 breast nodules smaller and larger than 20 mm

Status:

Version 1

Abstract

Figures

Background

Methods

Patients Selection

Instruments and strategies

Diagnostic Criteria

Statistical Analysis

Results

General information

Diagnostic results of BI-RADS and Co-Detect

Diagnostic performance of BI-RADS, S-Detect and Co-Detect

Discussion

Conclusions

Abbreviations

Declarations

References

Additional Declarations

Status:

Version 1