Applied Radiology

Mr. Grable is the CEO and director of Research and Development, Dr. Gkanatsios is Medical Physicist, and Dr. Ponder is the Advanced Development Manager with Imaging Diagnostic Systems, Inc., Plantation, FL.

The field of biomedical optics, the use of light in medicine, has seen rapid growth over the last decade. The development of non-invasive medical optical imaging (MOI) modalities using harmless near infrared (NIR) light is a major goal of biomedical optics research. Arguably, the most interesting application of MOI is breast imaging, with the potential to provide a safer, and possibly more effective, alternative to traditional x-ray mammography.

In 1929, Max Cutler reported the ability to differentiate between normal and pathological breast tissue by eye using an electric lamp to transilluminate the breast in a darkened room. ¹ Over the years, researchers improved upon this technique, known as diaphanography, by using more sensitive optical detectors and NIR light sources that have greater transmission through the breast. Typically, diaphanography images showed certain cancerous tumors as dark regions, thought to be caused by greater attenuation of NIR light by blood in the tumor-associated hypervascularisation. However, the highly scattering optical properties of breast tissues severely blurred the transmission images obtained with diaphanography. The limited spatial resolution, around 2 cm or worse, contributed to low sensitivity and specificity results during clinical trials of diaphanography in the 1980s preventing its acceptance as a clinically viable modality. ²

Technological advances in recent years, specifically pulsed lasers and ultra-fast detectors, have revived interest in MOI. The ability to illuminate the breast with picosecond pulses of NIR laser light and measure the temporal distribution of transmitted photons, known as the temporal point spread function (TPSF), enables investigation of time-resolved approaches to MOI. In contrast, diaphanography used a continuous wave (CW) light source and all the transmitted photons through the breast. Phantom studies have indicated that measuring the TPSF and using only the photons arriving earliest at a detector (a technique known as time-gating) can improve the image spatial resolution to about 1 cm. However, the poor signal-to-noise ratio obtained by measuring only the earliest photons ultimately limits the benefits of time-gating. Nevertheless, phantom studies have demonstrated that by fitting an analytic model based on diffusion theory to the TPSF and using the model's predictions for early photons the image spatial resolution of transmission images could be improved to about 5 mm. ³

An alternative to transmission images of the breast is to acquire optical images in a similar fashion to x-ray computed tomography (CT). Generally, this approach uses a modified back-projection algorithm that is analogous to x-ray CT, but incorporates the physics of NIR diffusion in the breast. ⁴ Back-projected slices yield the spatial variation in attenuation of the NIR light within the breast. A more sophisticated approach to reconstruct slice images uses a complex iterative inversion technique to predict TPSFs, or some characteristic of them, via a finite-element forward model based on diffusion theory. ⁵ This approach is theoretically more sound and enables the attenuation of NIR light to be de-coupled into an absorption image and a scattering image of the breast. It is anticipated this extra information will significantly help diagnosis.

Several universities and a growing number of companies are actively pursuing medical optical imaging with time-resolved methods. Some have chosen to pursue similar ideas in the frequency domain, which greatly reduces the complexity and expense of the imaging device. ⁶ At present, however, the frequency domain systems acquire less information content than time-resolved systems

Computer Tomography Laser Mammography

Since 1994, Imaging Diagnostic Systems Inc. (IDSI, Plantation, FL) has been developing a computed tomography laser mammography (CTLM ^™ ) system to serve as an adjunctive modality to x-ray mammography. The CTLM system operates as a tomographic scanner to generate slice images of the breast using non-ionizing NIR light and without applying any breast compression. The patient lies in a prone position with one breast pendent within the scanning chamber (figure 1). A bank of NIR-sensitive detectors, mounted around a horizontally rotating gantry, nearly encircles the breast without touching it. The breast is illuminated in the same horizontal plane mid-way between the first and last detectors by a pencil beam of NIR light from a laser source on the gantry (figure 2). A complete rotation of the gantry moves the pencil beam and bank of detectors around the breast. During rotation, the detectors measure the light transmitted through the breast from the pencil beam. Simultaneously, CCD cameras on the gantry view the pencil beam on the breast surface in order to measure the boundary shape of the breast. Once rotation is completed, the gantry can be moved vertically to another horizontal plane (slice) of the breast. During a typical scan, the CTLM system starts at the chest wall and acquires successive slices as the gantry is lowered down the breast. The acquired data is fed to an image reconstruction algorithm.

As mentioned previously, there are two main approaches to reconstructing slice images of the breast from optical CT data. Both a modified back-projection algorithm and an iterative inversion algorithm are being evaluated at IDSI. The reconstructed slice images can be displayed either volumetrically or as a series of coronal, sagittal, and axial profiles.

Throughout its continuing development, the CTLM system has existed as a variety of prototypes which all generally conform to the description outlined above. Both CW and time-resolved CTLM systems are in development; currently, the CW CTLM system is being evaluated in a clinical environment.

Preliminary Results

Phantoms with optical properties similar to breast tissue have been manufactured to help the development of the CTLM system. One of these phantoms is a breast-shaped epoxy-based shell with breast optical properties. The shell can be filled with a breast-like scattering solution and inclusions can be added to simulate tumors. For one experiment, the shell was filled with breast-like scattering solution and a small cylindrical inclusion (1 cm diameter, 1 cm height) was suspended inside it. The inclusion had the same transport scattering coefficient as the surrounding medium and 10 times greater absorption coefficient. The phantom was scanned using the CW CTLM system and slices were reconstructed with the modified back-projection algorithm. Figure 3 shows a volumetric image of the reconstructed slices of the phantom where the inclusion is clearly identified within the shell.

Phantom experiments have been good predictors of in vivo CTLM performance. Figure 4 shows results of a unilateral CTLM multislice scan of a normal volunteer with fatty breasts. Vascularisation and a well-localized area of dense tissue are visible.

A further avenue of development is the possibility of incorporating optical contrast agents to improve diagnosis. Fluorophores that absorb NIR light at one wavelength and fluoresce at another could be used as markers when injected into a patient. This technique could involve either specific targeting of the fluorophore to a certain type of tumor, or a non-specific approach relying on greater uptake of the fluorophore by the tumor due to hypervascularisation. Several major pharmaceutical companies are now expressing interest in developing optical contrast agents. These could be used in conjunction with a modified CTLM system, which has already demonstrated localization of a fluorescent inclusion within a breast-like phantom.

Future Direction

The results so far are highly encouraging and indicate the clinical potential of optical mammography. As expected, the spatial resolution shown in Figures 3 and 4 is poor since they were acquired with the CW CTLM system. However, it is anticipated that the spatial resolution will be greatly improved with the time-resolved CTLM system. Furthermore, the employment of optical contrast agents to aid diagnosis is an exciting proposition for the future. Ultimately, the outcome of clinical trials will determine the viability of the CTLM system in the breast-imaging market.

References

1. Cutler M: Transillumination as an aid in the diagnosis of breast lesions. Surg Gynecol Obstet 6:721-729, 1929.

2. Monsees B, Destouet JM, Gersell D: Light scan evaluation of nonpalpable breast lesions. Radiology 163:467-470, 1987.

3. Hebden JC, Hall DJ, Delpy DT: The spatial resolution performance of a time-resolved optical imaging system using temporal extrapolation. Med Phys 22:201-208, 1995.

4. Colak SB, Papaioannou DG, 't Hooft GW, et al: Tomographic image reconstruction from optical projections in light-diffusing media. Appl Opt 36(1):180-213, 1997.

5. Schweiger M, Arridge SR, Delpy DT: Application of the finite-element method for the forward and inverse models in optical tomography. J Math Imag Vis 3:263-283, 1993.

6. Fantini S, Francheschini MA, Gaida G, et al: Frequency-domain optical mammography: edge effect corrections. Med Phys 23:149-157, 1996.