The Automation of Breast Ultrasonography and the Medical Time Dedicated to the Method

In this integrative review, we aimed to describe the records of time devoted by physicians to breast ultrasound in a review of articles in the literature, in order to observe whether the automation of the method enabled a reduction in these values. We selected articles from the Latin American and Caribbean Literature in Health Sciences (LILACS) and MEDLINE databases, through Virtual Health Library (BVS), SciELO (Scientific Electronic Library Online), PubMed, and Scopus. We obtained 561 articles, and, after excluding duplicates and screening procedures, 9 were selected, whose main information related to the guiding question of the research was synthesized and analyzed. It was concluded that the automation of breast ultrasound represents a possible strategy for optimization of the medical time dedicated to the method, but this needs to be better evaluated in comparative studies between both methods (traditional and automated), with methodology directed to the specific investigation of this potentiality.


Introduction
The optimization of the medical workflow, while maintaining the accuracy of diagnostic methods, has been observed among the objectives of studies related to breast ultrasound. In its traditional form, breast ultrasound requires a medical time that is usually considered long. [1][2][3] In this context, automated breast ultrasound was developed, initially aiming at reducing the medical time for evaluating the ultrasound images, transferring the acquisition time of the same to a radiology technician, with specific training, allowing the use of the method on a large scale, for breast cancer screening. 1,4,5 The automated breast ultrasound device has a larger transducer than the conventional one, coupled to a mechanical arm, performing an automatic and standardized scan of the entire breast. The images obtained are transferred to a workstation where they are available for medical interpretation. 6,7 Three images are obtained (anteroposterior, lateral and medial of each breast), forming three planes or views for interpretation: coronal, sagittal, and transverse. 8,9 Factors such as the learning curve of the automated method, the physicians' experience with each of the methods, the number of findings, the size of the breasts (since a greater amount of breast tissue may require acquisition of additional views in the automated method and represents greater tissue volume to be evaluated also in the conventional method), interfere in this measure of time in an already established way. 3,8,10,11 The evaluation of the coronal view only, with the objective of reducing the time required for the physician to interpret the automated images, was analyzed by Schiaffino et al. Therefore, the multiplanar evaluation is mandatory, that is, all images must be obtained for a good diagnostic performance. 12 The use of computer algorithm systems to help detect changes in images obtained by automated ultrasound (computer-aided detection [CAD] system) is another strategy that has also been analyzed in some studies, with a reduction in medical interpretation time using these algorithms. 7,13 Thus, we aimed to describe the records of time dedicated by physicians to breast ultrasound in a review of literature articles, in order to observe whether the automation of the method made it possible to reduce these values.

Methods
This is an integrative literature review, developed observing the following steps: elaboration of the research question, selection of literature articles, data extraction and critical analysis of the included articles, presentation and discussion of the results obtained, and establishing the conclusion of the authors. 14 To define the question to be answered with the search for articles, the patient, intervention, comparison, and outcomes (PICO) strategy was used. 15 Our research object was the medical time required for breast evaluation using the automated way of obtaining the images. The intervention was defined as the use of the automated method of ultrasound of the breasts and our comparison was established with the conventional method of performing this exam, with the expectation as an outcome to reduce this medical time with the use of the automated method. Thus, we used the following question to guide our review: How long does the physician need to evaluate the automated ultrasound images of the breasts? Would this time be shorter than the time required to perform a conventional (non-automated) ultrasound of the breasts?
The selection of articles was made in July and August of 2022 in the Latin American and Caribbean Literature in Health Sciences (LILACS) and MEDLINE databases, through the Virtual Health Library (BVS), Scientific Electronic Library Online (SciELO), PubMed, and Scopus. As descriptors, in Health Sciences (DeCS) and Medical Subject Headings (MeSH), we used mammary ultrasonography, breast ultrasonography, diagnostic imaging, breast neoplasms, and threedimensional imaging.
We applied language filters, selecting articles in English and Portuguese, with full text available, and selected screening, diagnosis, prognosis, evaluation, and observational studies in the areas of medicine, imaging, gynecology, and radiology as the type of studies.

Results
We obtained 561 articles from the databases, and, using the Rayyan application (Qatar Computing Research Institute, Ar-Rayyan, Qatar) 16 , 45 duplicate articles were found, leaving 516 articles for analysis. Of these, 453 were excluded and 63 were included by reading the title. Of the 63 included, 22 were excluded, and 41 were included after reading the abstract. These 41 included articles were then considered for full text reading. After reading the full text, 32 were excluded, 12 of which did not present the measurement of the medical time spent interpreting the images obtained by automated breast ultrasound (reason 1), 10 in relation to the time to perform the conventional breast ultrasound (reason 2) and 6 for both methods (reason 3), and 4 for being narrative review articles (reason 4). The remaining 9 articles provided data for the composition of ►Charts 1, 2, and 3. 1,2,10,11,13,17-19,21 ►Figure 1 summarizes these results in the PRISMA 2020 flowchart. 22

Discussion
Considering the guiding question of this review, the medical time dedicated to the two methods of breast evaluation by ultrasound, we observed with the data from the studies included in this review that less medical time was spent on the automated method in most studies, but with few studies directly comparing both methods regarding the specific question of medical time dedicated to each one of them. 1,2,10,11,13,17,19,21,23 Of the nine selected studies, seven brought only time information for one of the methods, either because the measurement of this time had not been included in the methodology of these studies or because the comparison Chart   The two studies that presented the time for both methods differed in their conclusions regarding medical time. 11,18 Tutar et al. included 340 patients in a cross-sectional study in which the average time for interpretation of automated ultrasound images was 14.5 minutes, greater than the average of 12.5 minutes observed for conventional ultrasound, with data reported descriptively. The authors attributed this result to the fact that they recorded all the lesions observed and analyzed all the images of the coronal, transverse, and longitudinal planes of each of the views (anteroposterior, lateral, and medial) obtained for each of the breasts in the automated ultrasound. 18 However, a similar analysis was cited in the methodology of studies that measured medical time for interpretation of automated images. 10,13,17,21,23 The study by Skaane et al. stands out, with results that reinforce the observation that the number of findings interferes with the time required for image analysis. For the analysis of the images of both breasts, they obtained, on average, 9 minutes, and, considering the time of each breast individually, normal breasts or breasts with cysts required an average of 4 minutes, while breasts with probably benign nodules required 4.8 minutes, and breasts with suspicious findings for cancer required an average of 5.3 minutes. 10 The other study that uses time data for both methods also has a cross-sectional design, including 440 patients. This study brings in its methodology the particularity of the different time of execution of conventional ultrasound by breast radiologists (average time of 7 minutes and 45 seconds) and by radiologists not specialized in breast imaging (average time of 4 minutes and 15 seconds). Automated ultrasound data were interpreted only by breast radiologists, in an average time of 4 minutes and 25 seconds. The difference between the means of the breast radiologists was analyzed for both methods using the t-Student test and was considered statistically significant (p < 0.001). 11 Philadelpho et al. (2021) and Tutar et al. (2020) included patients with dense breasts in breast cancer screening in their studies. High-risk patients and those who had already been diagnosed and were being followed up were excluded, thus sampling a population whose exams tend to present fewer findings. Therefore, Philadelpho et al.  Thus, we observed that even considering only the time variable, many factors are associated and interfere with its measurement, probably explaining the variability of data obtained in the literature for both conventional and automated methods of breast ultrasound̀evaluation.
As a limitation of this review, we have the small number of studies that evaluated the medical time in both methods, the fact that they are studies with a lower level of evidence, level 4, and the question that only one of them included a statistical analysis of the difference between the averages obtained for the time variable.
These observations suggest that the comparison of the times spent by the physician with each of the methods needs to be better evaluated in experimental studies, with a larger number of patients, which could allow a better evaluation of the potential of automated ultrasound in optimizing medical time.

Conclusion
In our integrative literature review, the automation of breast ultrasound represents a possible strategy for optimizing the medical time dedicated to the method, but it needs to be better evaluated in comparative studies between both methods, with a methodology aimed at the specific investigation of this potentiality.