Prospective evaluation of breast lesions categorisation with S-detect: feasibility and inter-observer variability assessment

Congress:

ECR 2018

Poster Number:

C-1228

Type:

Scientific Exhibit

Keywords:

Breast, Ultrasound, CAD, Computer Applications-Detection, diagnosis

Authors:

V. De Soccio, V. Cantisani, G. Bonito, V. Magri, M. Di Segni, A. Rubini, C. De Felice, F. D'Ambrosio, C. Catalano; Rome/IT

DOI:

10.1594/ecr2018/C-1228

DOI-Link:

https://dx.doi.org/10.1594/ecr2018/C-1228

Fig. 1: Radiologist with 32 years experience in breast imaging

Fig. 2: Radiologist with 18 years exoerience

Fig. 3: 5th year with limited experience in breast imaging

Fig. 4: 2 years with deeper experience in breast imaging

Aims and objectives

US has a pivotal role in breast imaging especially for breast lesions characterization. Furthermore, to improve the performance of US, other tools were developed, such as elastography, Computer-Aided Diagnosis (CAD) systems [1] and S-detect. S-detect is a software based upon the deep learning algorythm, which performs lesion segmentation, feature analysis and describes, according to either BIRADS 2003 or BIRADS 2013, lexicon and suggests a dichotomic categorization. In addition to the role as a possible adjunct tool in breast lesion characterization, another possible application could be...

Methods and materials

Patient population: The study design was prospective. Between July 2016 and June 2017, 122 female patients aged between 21 and 84 years (mean 51 years) underwent baseline US examination with linear array probes. The following inclusion criteria were used: Patients with follow-up in progress or who previously underwent cytology or biopsy sampling due to likely benign lesions; Patients candidate to biopsy due to lesions suspicious for malignancy. The exclusion criteria employed were: Pregnancy; Lactation; Neo-adjuvant chemotherapy in progress or at less than 2 months since...

Results

Breast lesions: 68 nodules in 61 patients were analyzed, with size between 10 and 48 mm; of these nodules, 44 were malignant and 24 were benign. Among 44 malignancies, 37 were IDC, 3 were DCIS, 3 were Infiltrating Lobular Carcinomas and 1 was a Granular Cell Tumor. Among 24 benign lesions there were: 12 were fibroadenomas (one of them was characterized through a biopsy and one was characterized with cytology and was being followed-up), 1 Phyllodes tumor, 2 hamartomas (characterized through the sonographic appearance and...

Conclusion

Development and publication of BI-RADS began after the recognition of the need of a globally shared lexicon that could allow sharing and clear expression of morphology, operator’s judgment and of the strategy considered to be the most advised in the assessment of breast lesions [2]. Compared to the former edition, the latest includes some changes, including special cases introduction, changes in the description of surrounding tissues, of calcification and vascularity [5]. According to Xiaoyun Xiao et al [6], the use of 2013 criteria showed 100%...

Personal information

G.Alagna, MD Università Sapienza di Roma,Policlinico Umberto I,Department of Radiology-Italy. Viale del Policlinico, 155 00161 Roma RM, Italy Phone: + 39 328 3745751 e-mail: [email protected]

References

Jalalian A, Mashohor SB, Mahmud HR, Saripan MI, Ramli AR, Karasfi B (2013) Computer-aided detection/diagnosis of breast cancer in mammography and ultrasound: a review. Clin Imaging; 37:420-426. 2. D’Orsi CJ, Sickles EA, Mendelson EB, Morris EA et al. (2013) ACR BI-RADS® Atlas, Breast Imaging Reporting and Data System. Reston,VA, American College of Radiology. Jales RM, Sarian LO, Torresan R, Marussi EF, Alvares BR, Derchain S (2013) Simple rules for ultrasonographic subcategorization of BI-RADS®-US 4 breast masses Eur J Radiol; Aug;82(8):1231-5. Landis JR, Koch GG (1977)...