Photoacoustic imaging has the potential for real-time molecular imaging at high resolution and deep inside the tissue, using non-ionizing radiation and not necessarily depending on exogenous imaging agents, making this technique very promising for a range of clinical applications. The fact that photoacoustic imaging systems can be made portable and compatible with existing imaging technologies favors clinical translation even more. The breadth of clinical applications in which photoacoustics could play a valuable role include: noninvasive imaging of the breast, sentinel lymph nodes, skin, thyroid, eye, prostate (transrectal), and ovaries (transvaginal); minimally invasive endoscopic imaging of gastrointestinal tract, bladder, and circulating tumor cells (in vivo flow cytometry); and intraoperative imaging for assessment of tumor margins and (lymph node) metastases. In this review we describe the basics of photoacoustic imaging and its recent advances in biomedical research, followed by a discussion of strategies for clinical translation of the technique.
Background
Artificial intelligence (AI) systems performing at radiologist-like levels in the evaluation of digital mammography (DM) would improve breast cancer screening accuracy and efficiency. We aimed to compare the stand-alone performance of an AI system to that of radiologists in detecting breast cancer in DM.
Methods
Nine multi-reader, multi-case study datasets previously used for different research purposes in seven countries were collected. Each dataset consisted of DM exams acquired with systems from four different vendors, multiple radiologists’ assessments per exam, and ground truth verified by histopathological analysis or follow-up, yielding a total of 2652 exams (653 malignant) and interpretations by 101 radiologists (28 296 independent interpretations). An AI system analyzed these exams yielding a level of suspicion of cancer present between 1 and 10. The detection performance between the radiologists and the AI system was compared using a noninferiority null hypothesis at a margin of 0.05.
Results
The performance of the AI system was statistically noninferior to that of the average of the 101 radiologists. The AI system had a 0.840 (95% confidence interval [CI] = 0.820 to 0.860) area under the ROC curve and the average of the radiologists was 0.814 (95% CI = 0.787 to 0.841) (difference 95% CI = −0.003 to 0.055). The AI system had an AUC higher than 61.4% of the radiologists.
Conclusions
The evaluated AI system achieved a cancer detection accuracy comparable to an average breast radiologist in this retrospective setting. Although promising, the performance and impact of such a system in a screening setting needs further investigation.
Objective To evaluate the rate of over-diagnosis of breast cancer 15 years after the end of the Malmö mammographic screening trial. Design Follow-up study. Setting Malmö, Sweden. Subjects 42 283 women aged 45-69 years at randomisation. Interventions Screening for breast cancer with mammography or not (controls). Screening was offered at the end of the randomisation design to both groups aged 45-54 at randomisation but not to groups aged 55-69 at randomisation. Main outcome measures Rate of over-diagnosis of breast cancer (in situ and invasive), calculated as incidence in the invited and control groups, during period of randomised design (period 1), during period after randomised design ended (period 2), and at end of follow-up. Results In women aged 55-69 years at randomisation the relative rates of over-diagnosis of breast cancer (95% confidence intervals) were 1.32 (1.14 to 1.53) for period 1, 0.92 (0.79 to 1.06) for period 2, and 1.10 (0.99 to 1.22) at the end of follow-up. Conclusion Conclusions on over-diagnosis of breast cancer in the Malmö mammographic screening trial can be drawn mainly for women aged 55-69 years at randomisation whose control groups were never screened. Fifteen years after the trial ended the rate of over-diagnosis of breast cancer was 10% in this age group.
The main purpose was to compare breast cancer visibility in one-view breast tomosynthesis (BT) to cancer visibility in one- or two-view digital mammography (DM). Thirty-six patients were selected on the basis of subtle signs of breast cancer on DM. One-view BT was performed with the same compression angle as the DM image in which the finding was least/not visible. On BT, 25 projections images were acquired over an angular range of 50 degrees, with double the dose of one-view DM. Two expert breast imagers classified one- and two-view DM, and BT findings for cancer visibility and BIRADS cancer probability in a non-blinded consensus study. Forty breast cancers were found in 37 breasts. The cancers were rated more visible on BT compared to one-view and two-view DM in 22 and 11 cases, respectively, (p < 0.01 for both comparisons). Comparing one-view DM to one-view BT, 21 patients were upgraded on BIRADS classification (p < 0.01). Comparing two-view DM to one-view BT, 12 patients were upgraded on BIRADS classification (p < 0.01). The results indicate that the cancer visibility on BT is superior to DM, which suggests that BT may have a higher sensitivity for breast cancer detection.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.