Ultrasound-DOT Reduced Benign Biopsy Recommendations in Study

A hand-held imaging system that combines ultrasound with diffuse optical tomography, or DOT, reduced radiologists' recommendations to biopsy benign breast lesions in a prospective study. The result points to a possible way to add functional information to conventional breast ultrasound, but it came with several incorrect downgrades of cancer readings.

The study enrolled women who had ultrasound-visible abnormalities and were already scheduled for an ultrasound-guided core needle biopsy. All participants still underwent biopsy, allowing the imaging assessments to be compared with tissue pathology. The work therefore tested a diagnostic aid under controlled conditions; it did not show how many procedures would safely be avoided in routine practice.

The distinction is important because breast biopsy remains the definitive method used to determine whether a suspicious lesion is malignant. The proposed technology is meant to help radiologists decide which lower-suspicion lesions might instead be monitored. It is not designed to replace mammography, ultrasound or pathology.

DOT Adds Blood and Oxygen Measurements to Ultrasound

Conventional ultrasound provides structural information about a breast abnormality. It can usually distinguish a fluid-filled cyst from a solid mass, but benign and malignant solid lesions can share visible features. Suspicious findings are assigned a Breast Imaging Reporting and Data System, or BI-RADS, category that guides whether biopsy or imaging follow-up is recommended.

Diffuse optical tomography adds a different type of information. The system directs near-infrared light into breast tissue and measures reflected light at four wavelengths. Those data are used to estimate total hemoglobin concentration and blood oxygen saturation in and around the lesion. Tumors can develop abnormal blood vessels and altered oxygen use, so these functional measurements may help distinguish some cancers from benign tissue.

The research device placed a conventional ultrasound transducer inside a larger hand-held DOT probe. Ultrasound located the lesion, while measurements from the corresponding area of the opposite breast provided a reference for optical reconstruction. Image acquisition took about five to ten minutes, and data processing and reconstruction took about 20 minutes.

This is diffuse optical tomography, not photoacoustic imaging. The two technologies can measure related vascular features but use different hardware and reconstruction methods. The study did not test an artificial-intelligence classifier as part of the reported clinical reading process, although the researchers are developing machine-learning tools for future versions.

How the 226-Patient Study Was Run

Researchers recruited 238 women at a single academic center between July 2020 and August 2024. Twelve were excluded from analysis, leaving 226 participants with a mean age of 52 and an age range of 19 to 85. Pathology identified 70 invasive carcinomas, seven cases of ductal carcinoma in situ, nine high-risk lesions and 140 benign lesions.

Three or four breast radiologists independently reviewed each case. They first assigned BI-RADS assessments using conventional ultrasound and mammography when available. They then reviewed the DOT maps, total hemoglobin and oxygen-saturation measurements before assigning a second assessment. The readers, engineers and study personnel were blinded to biopsy outcomes during those decisions.

The analysis contained 823 reader assessments because several radiologists interpreted each of the 226 cases. DOT information did not change the biopsy recommendation in 630 readings. Among 485 readings of benign lesions, 119 were appropriately downgraded to a category that would favor imaging follow-up instead of immediate biopsy. That is the reported 24.54% reduction in benign biopsy recommendations.

The figure describes reader-level recommendations, not 25% of enrolled patients and not a measured reduction in procedures. Everyone in the study had a biopsy so researchers could establish the final diagnosis. A real-world trial would need to show that patients assigned to follow-up can be monitored safely and that cancers are not diagnosed at a more advanced stage.

The False Negatives Define the Safety Question

Adding DOT data also caused five of 281 malignant-lesion readings to be incorrectly downgraded, a false-negative rate of 1.78%. Those five readings involved three patients. Three came from different readers assessing a lactating patient whose breast physiology suppressed the expected hemoglobin signal. Another involved an invasive lobular carcinoma initially interpreted in the context of recent trauma, and the fifth involved a small ductal carcinoma in situ.

The researchers compare the 1.78% reading-level rate with the BI-RADS 3 category, which carries an expected malignancy risk of no more than 2% and ordinarily leads to short-interval follow-up rather than immediate biopsy. That comparison does not make a missed cancer harmless. It identifies the safety threshold a diagnostic adjunct would have to meet while ensuring that recommended follow-up actually occurs.

Total hemoglobin was significantly higher and oxygen saturation significantly lower in the combined malignant group than in the combined benign and high-risk group. The measurements alone were not accurate enough to guide decisions: a total-hemoglobin threshold of 70 micromolar produced 71.4% sensitivity and 68.5% specificity, while an oxygen-saturation threshold of 65% produced 65.8% sensitivity and 51.1% specificity.

The system's value in this experiment came from combining DOT data with radiologists' interpretation of conventional imaging. Agreement among readers remained moderate, and agreement after DOT was not better than with conventional imaging alone. A standardized scoring framework and training process would be needed before results could be reproduced across hospitals.

Promising Triage Evidence Is Not Clinical Adoption

The study has limits that directly affect where the result can be used. It was conducted at one academic institution with specialist breast radiologists. Only ultrasound-visible abnormalities already heading to biopsy were included. Some superficial lesions, findings beneath the nipple and areas with a recent biopsy were excluded because they could interfere with DOT acquisition.

The technology also required measurements from the opposite breast and around 20 minutes of processing before interpretation. It has not been validated across community sites, general radiologists, different equipment or a population arriving for routine screening. The study did not measure clinical implementation or national spending.

A credible next step would be a multicenter study with a locked interpretation protocol and patient-level outcomes. It should report how many biopsies were deferred, whether follow-up was completed, which cancers were delayed or missed, and how performance changed during lactation or with lesion location and cancer subtype.

The present result is useful because it measures both sides of the tradeoff. DOT information reduced a substantial share of benign-lesion biopsy recommendations, while the pathology results exposed the cancers that could have been incorrectly monitored. Whether the balance is safe enough for clinical use depends on replication and follow-up evidence, not on treating a 25% reader-level reduction as a finished diagnostic solution.