Breast cancer is the most frequent female malignancy in developed countries. In Italy almost 40,000 new cases were diagnosed in 2010, with more than 7,000 deaths [
The use of transillumination of the breast dates back to the 1920s, when it was proposed to investigate breast cancer [
The dynamic optical breast Imaging (DOBI) has recently shown to be effective in detecting breast cancer in young women [
In a recent study the DOBI ComfortScan was evaluated in 46 patients with Breast Imaging Reporting and Data System (BI-RADS) 3–5 breast lesions. The DOBI ComfortScan examination showed statistical difference in number of suspect pixels between benign (
The present multicenter study aimed at increasing the diagnostic accuracy of DOBI ComfortScan in a larger population and at assessing whether it can be a useful tool for breast cancer screening.
We retrospectively reviewed the medical records of patients who had undergone surgical excision of suspicious or positive breast lesions at the National Cancer Institute of Naples and at the Clinical Zucchi of Milan between March 2009 and July 2011. All patients had been considered eligible for surgery since they showed BI-RADS 4-5 breast lesions at standard imaging (mammography and breast ultrasound). Patient’s workout included medical history, breast abdominal and pelvic ultrasound, mammography, chest X-ray, and routine blood chemistry with tumor markers assessment (CA 15.2 and CEA). Cytological or microhistological characterization of the lesions was also acquired. Breast magnetic resonance imaging (MRI) was performed in case of suspicious multicentric and/or bilateral breast cancer. Bone scan was also included in the staging in case of malignancy. All patients who underwent Dynamic Optical Breast Imaging in the preoperative workflow were considered eligible for the study. Four patients were excluded due to insufficient illumination of the breasts. The study was approved by the local ethics committees and was performed in accordance with the current version of the Declaration of Helsinki and the International Conference on Harmonization of Good Clinical Practice Guidelines. Informed consent was obtained from all patients after the nature of the examinations was explained fully.
The DOBI Medical ComfortScan system is an advanced digital imaging device that uses high-intensity, light-emitting diodes (LEDs) and gentle external pressure to highlight areas with vascular abnormalities. The high-intensity LEDs transmit red light through the breast, if the light encounters a neoangiogenic region, it is absorbed or scattered differently than in other regions of the breast. This is caused by the different concentrations of oxygenated and deoxygenated hemoglobin.
The system is made of three physical assemblies (Figure
The patients stand in front of the machine and the breast is positioned onto the panel of the C-arm assembly. The LEDs of the system illuminated the breast and the light (with wavelength of 640 nm) is transmitted through the tissues and quantified on the other side by the charge coupled device (CCD) camera. In the acquisition window of the operating software, the operator can mark the region of interest (ROI) placing two pointers indicating, respectively, the nipple and where it is supposed to be the lesion.
A soft transparent silicone membrane is placed in contact with the upper surface of the breast and then inflated under computer control during the exam. The pressure is set to 5 mmHg for the first 15 s of the scan, raised to 10 mmHg over the next 30 s (the dynamic image sequence), and decreased back to 5 mmHg for the final 15 s. Forty-five frames are overall acquired (five baseline before applying pressure and 40 during the dynamic sequence). The transmitted light is detected by the CCD throughout the scan and recorded by the computer that processed the data to generate the dynamic angiogenic signature (DAS), a sequence of images of the breast in a cranial-caudal view.
The analysis of the Dynamic Angiogenic Signature may identify changes in local blood perfusion and oxygen saturation as variations in image contrast defined area of pathologic interest (API). The API presents an increase of both blood volume and depletion of blood oxygen with a reduction of the amount of light reaching the CCD camera. This feature appears like an area of decreased intensity (dark blue) or highly decreased intensity (purple) of color contrast. The ComfortView software selects the optimal contrast “auto contrast” for display of the study varying the scale (percentage) used along the
We defined twelve items to correctly estimate the API in order to standardize the criteria of the exam interpretation. The parameters were blindly evaluated and grouped into three categories: morphologic, spatial, and temporal items (Table
Criteria for exam interpretation.
Class of item | Characteristic of API1 | Description | Figure |
---|---|---|---|
Morphologic | Margins | Absence of API1 | Figure |
Shaded | Figure |
||
Well defined | Figure |
||
Margins in zoom mode | Absence of API or margins indistinct | ||
Shared or well defined | |||
Major diameter | Absence of API1 or API > 5 cm | ||
From >3 to 5 cm | |||
From 1 to 3 cm | |||
| |||
Spatial | Location of API1 in the ROI2 marker | Absence of API or out of the rings | |
Within the third ring | |||
Within the second ring | |||
Within the first ring | |||
Distance of the API from the nipple marker | ≤2 cm from the nipple | ||
>2 cm from the nipple | |||
Maximum pixel intensity | <2050 | ||
From 2050 to 4000 | |||
>4000 | |||
| |||
Temporal | Time of API detection | Absence of API | |
From 10 to 20 seconds | |||
<10 seconds | |||
Duration of API at the dynamic analysis | Absence of API or <10 seconds duration | ||
Duration from 10 to 20 seconds | |||
Duration >20 seconds | |||
Temporal signature graft in auto mode | Absence of API | Figure |
|
Curve with upward trend | Figure |
||
Sinusoidal curve | Figure |
||
Curve slowly downward | Figure |
||
Curve rapidly downward | Figure |
||
Linear curve rapidly downward | Figure |
||
Temporal signature graft in zoom mode | Absence of API | ||
Curve with upward trend | |||
Sinusoidal curve | |||
Curve slowly downward | |||
Curve rapidly downward | |||
Linear curve rapidly downward | |||
Similar/dissimilar temporal signature grafts in auto mode | Absence of API or curves very similar | Figure |
|
Curves slightly different | Figure |
||
Curves very dissimilar | Figure |
||
Similar/dissimilar temporal signature grafts in zoom mode | Absence of API or curves very similar | ||
Curves slightly different | |||
Curves very dissimilar |
1API: area of pathologic interest; 2ROI: region of interest.
Morphologic features of the area of pathologic interest (API).
Margins of the API not evaluable
Margins of the API shaded but evaluable
Margins of the API well defined
The spatial items define the distance of the API from the nipple marker and its location in the ROI marker. The lesion is more relevant when enclosed in the ROI marker and localized distant from the nipple to avoid the “nipple blue-phenomenal” due to the natural convergence of the vessels to the nipple. The maximum pixel intensity of the API also correlates with malignancy. An intensity of −4000 is considered predictive of malignancy, while −2050 to −4000 is less strongly predictive, but still likely malignant. An intensity of 0 to −2050 is considered indeterminate (consistent with a “flat” or nearly flat curve).
The temporal items evaluate the time of API detection and its duration and the temporal signature graft (in auto and zoom mode), which represents the changes over time of the pixel intensity as well as the differences of the temporal signature grafts between the API and the normal breast tissue in auto and zoom mode. The absence of API means absence of detectable neangiogenesis (Figure
Temporal signature graft curves. The blue line indicates the temporal signature graft curves. The pressure graph is displayed under the temporal signature graph (Yellow line).
Absence of the area of pathologic interest (API)
Upword trend of the curve
Sinusoidal trend of the curve
Curve slowly downward
Curve rapidly downward
Linear curve rapidly downward
Curve becoming more positive over time (trend is upward) indicates that less light is being absorbed and is consistent with a benign area. This would appear as an area of green that becomes orange/yellow/white over the sequence of images (Figure
If the curve is becoming more negative over time (the trend is downward), this indicates that more light is being absorbed and is consistent with the presence of malignancy. A more rapid decrease of the curve is more likely associated with malignancy (Figures
Comparison between temporal signatures of the API (blue curve) with a control marker placed on a non-pathologic area of the breast (red curve).
Curves showing a very similar shape
Curves not distinctly different
Curves showing a very dissimilar shape
We scored the 12 variables using a standardized code (Table
The DOBI score derived from the discriminant analysis.
Class of item | Characteristic of API1 | Code | Description | Figure |
---|---|---|---|---|
Morphologic | Margins | 0 | Absence of API1 | Figure |
1 | Shaded | Figure |
||
2 | Well defined | Figure |
||
Margins in zoom mode | 0 | Absence of API or margins indistinct | ||
1 | Shaded or well defined | |||
Major diameter | 0 | Absence of API1 or API > 5 cm | ||
1 | From >3 to 5 cm | |||
3 | From 1 to 3 cm | |||
| ||||
Spatial | Maximum pixel intensity | 0 | <2050 | |
2 | From 2050 to 4000 | |||
4 | >4000 | |||
| ||||
Temporal |
Time of API detection | 0 | Absence of API | |
1 | From 10 to 20 seconds | |||
2 | <10 seconds | |||
Duration of AIP at the dynamic analysis | 0 | Absence of API or <10 seconds duration | ||
1 | Duration from 10 to 20 seconds | |||
2 | Duration >20 seconds | |||
Temporal signature graft in auto mode | 0 | Absence of API | Figure |
|
1 | Curve with upward trend | Figure |
||
2 | Sinusoidal curve | Figure |
||
3 | Curve slowly downward | Figure |
||
6 | Curve rapidly downward | Figure |
||
9 | Linear curve rapidly downward | Figure |
||
Temporal signature graft in zoom mode | 0 | Absence of API | ||
1 | Curve with upward trend | |||
2 | Sinusoidal curve | |||
3 | Curve slowly downward | |||
6 | Curve rapidly downward | |||
9 | Linear curve rapidly downward | |||
Similar/dissimilar temporal signature grafts in auto mode | 0 | Absence of API or curves very similar | Figure |
|
3 | Curves slightly different | Figure |
||
6 | Curves very dissimilar | Figure |
||
Similar/dissimilar temporal signature grafts in zoom mode | 0 | Absence of API or curves very similar | ||
1 | Curves slightly different | |||
2 | Curves very dissimilar |
DA was used to model the value of a dependent categorical variable based on its relationship to one or more predictors. We calculated the Wilks’ lambda for each variable to define its correlation with histology (benign versus malignant). The variables that resulted significantly associated with histology at discriminant analysis were recoded with a new score according to the strength of their correlation.
The summation score derived from the discriminant analysis was evaluated by using the receiver-operating characteristic (ROC) curves to assess the best cutoff. The area under ROC curve (AUC) illustrates the performance of a classifier over all sensitivity and specificity levels. The AUC can range from 0.5 (chance performance) to 1.0 (perfect performance). In order to minimize the rate of false negative, we decided to take into account a performance at very high sensitivity level.
Finally we developed a 4-class DOBI-level correlating with pathologic findings: malignant (category B5); possibly malignant (category B4); benign but the possibility of malignancy cannot be excluded (category B3); benign (category B2).
The chi-square test was used to assess the relationship between the DOBI score and prognostic variables.
All of the statistical analyses were performed using SPSS (version 16.0).
Overall 113 women underwent DOBI-ComfortScan before surgery. Median age was 49.8 (range 25–79) years. Twenty-seven radical mastectomies, 47 quadrantectomies, and 39 wide excisions were performed in the 113 patients reviewed. Overall, 65 invasive cancer, 9 in situ carcinoma and 39 nonmalignant lesions, were observed. Fifty-three patients out of 74 patients with carcinoma presented positive axillary nodes. Pathological tumor size was T1, T2, and T3 in 41, 20, and 4 patients, respectively. Histotype of patients with invasive carcinoma was as follows: ductal carcinoma 54, lobular carcinoma 6, others 5. At immunochemistry, 54 patients had HR positive, 16 patients presented an HER2 overexpressing, and 4 cases were triple negative tumor Table
Discriminant analysis of the DOBI Scan results was validated originating a classification matrix with 80% of original grouped cases correctly classified and 85.0% of cross-validated grouped cases correctly classified, having confirmed a major contribution of the same variables that had significant correlation coefficients.
The relationship between the categories and score of various items is shown in Table
Association between DOBI score and some prognostic factors.
DOBI score | |||||
---|---|---|---|---|---|
Patient characteristics | <30,5 |
≥30,5 |
|
||
No. |
|
No. | % | ||
Total | |||||
Age | 0.08 | ||||
≤40 | 12 | 24.5 | 7 | 10.9 | |
40–49 | 21 | 42.9 | 22 | 34.4 | |
50–59 | 8 | 16.3 | 16 | 250 | |
≥60 | 8 | 16.3 | 19 | 29.7 | |
| |||||
Tumor size | 0.7 | ||||
T1 | 8 | 57.1 | 32 | 61.5 | |
T2-T3 | 6 | 42.9 | 20 | 38.5 | |
| |||||
Grading | 0.5 | ||||
I | 0 | 4 | 7.4 | ||
II | 7 | 50.0 | 26 | 48.1 | |
III | 7 | 50.0 | 24 | 44.4 | |
| |||||
Cerb2 | 0.3 | ||||
Positive | 1 | 6.7 | 9 | 15.8 | |
Negative | 14 | 93.3 | 48 | 84.2 | |
| |||||
Receptor status | 0.2 | ||||
Positive | 9 | 64.3 | 45 | 80.4 | |
Negative | 5 | 35.7 | 11 | 19.6 |
The following variables better discriminated the function: major diameter; pixel intensity; temporal signature graft in auto mode; temporal signature graft in zoom mode.
The summation score, made up of 10 variables (two were excluded since they resulted not significant), was evaluated by using the receiver operating characteristic (ROC) curves to assess the best cutoff; these are plots of the true positive rate against the false positive rate for different possible cut-off points (Figure
The Roc curve.
The DOBI ComfortScan System device.
Finally, the DOBI score was related to diagnostic results (DOBI Level) such as malignant (category B5 ≥ 38 score); possibly malignant (category B4 = 25–37 score); benign but the possibility of malignancy cannot be excluded (category B3 = 20–24 score); benign (category B2 < 20 score).
Table
The DOBI-level.
Level | Score |
|
Cancer | No cancer | PPV** | NPV*** |
---|---|---|---|---|---|---|
1 |
|
0 | 0 | 0 | 0 |
|
2 | <20 | 10 | 0 |
|
0 |
|
3 | 20–24 | 14 | 3 | 11 |
|
|
4 | 25–37 | 56 | 39 | 17 |
|
|
5 |
|
33 | 32 | 1 |
|
|
Breast cancer is the first cause of cancer-related death among women in western countries. In 2010, the standardized mortality rate was estimated to be 13.3/100.000 worldwide and 30.9/100.000 in developed countries [
The purpose of our study was to improve the accuracy of DOBI ComfortScan by evaluating other parameters besides maximum pixel intensity. We considered 12 items, grouped in three categories, spatial, temporal, and localizing since we hypothesized that the temporal changes of the pixel intensity as well as the spatial characteristics of the API and its relationship with normal breast could add useful information to achieve a correct diagnosis. The discriminant analysis permitted to exclude two spatial items, the location of the API in the ROI marker and its distance from the nipple. The remaining ten items, which had significantly correlated with histology, were recoded according to strength of their correlation. A higher score was given to the following five parameters which better discriminated the function: major diameter, pixel intensity, and temporal signature graft in auto and zoom mode as well as the correlation between the temporal signature grafts of API and normal tissue. The ROC analysis showed that the best cutoff of DOBI score was 30.5, which resulted in a sensitivity and specificity of 80% and 87%, respectively. A DOBI score > 25 resulted in a much higher sensitivity (93%) and a DOBI score > 21 reached a 99% sensitivity although the specificity dramatically fell. According to the performance depicted in the ROC curve, we developed a DOBI 0–5 categorical classification (DOBI-level, Table
The DOBI ComfortScan is a low-cost, noninvasive technique with a good potential for discriminating benign from malignant lesions. According to our results, the definition of other parameters besides pixel intensity permits to improve the accuracy of this diagnostic procedure. Further studies are warranted to define the potential role of DOBIComfortScan in breast cancer imaging.