"We believe that THz imaging could be used as a fast screening technique to indicate where specific regions of interest are in the sample," researcher Florian Formanek from Sony's Life Science Laboratory told medicalphysicsweb. "A THz image could produce a global view of a tissue section based on microscopic information, thus reducing the need for doctors to investigate numerous parts at high magnification with an optical microscope. It may also reduce the diagnosis dependence on the doctor's experience."
Sitting between the far-infrared and microwave bands, terahertz waves are a form of non-ionizing radiation that can penetrate a wide variety of materials, such as clothing, plastics and ceramics – but notably, not water. Knowing that THz radiation can identify cancerous tissue, the Sony researchers teamed up with the Tokyo Medical and Dental University to develop a tool that would assist pathologists with their diagnosis.
The imaging system
The first step was to build a THz imaging system working in reflection geometry. THz radiation is generated using photoconductive antennas gated by femtosecond pulses from a Ti:sapphire laser. The sample is raster scanned through the focused THz beam and the reflections are collected by a detector. A typical image consists of 100 × 100 pixels, where each pixel corresponds to a time-domain spectrum of 1024 points.
Formanek and his colleagues studied two samples. "We used 10 µm thick sections cut from paraffin-embedded tissue blocks," commented Formanek. "These samples are readily available from the archive of any pathology department, thus allowing for a large scale study and direct registration with microscope images for contrast investigation and database building."
The two samples under investigation were a squamous cell carcinoma of the lung and a ductal carcinoma of the pancreas. In both cases, a visible image was acquired simultaneously with the THz image. Data gathered by the THz imaging system in the time domain was Fourier transformed into the frequency domain for analysis.
Image analysis
Image analysis involved obtaining the complex refractive index at each pixel and producing a segmented image by applying a clustering algorithm to the data. "A segmented image clearly identifies similar areas," explained Formanek. "The clustering algorithm allows us to group pixels with a similar refractive index, producing an image where tissues having the same spectroscopic properties in the THz frequency range can be attributed the same label or colour."
Thorough THz investigation of the lung-cancer sample revealed two tissue subtypes within the tumour, two distinct types of healthy tissues and a necrotic area close to the centre of the tumour. "Our measurements show that a low magnification visible image does not have a clear distinction between the various areas within the tumour," commented Formanek. "Inspection of visible images using a high magnification confirms that the segmentation obtained on the THz image corresponds to differences of cellular type at the microscopic level."
In the case of the pancreatic cancer sample, the tumour was so invasive that a highly magnified visible image was required to distinguish cancerous and healthy cells. In comparison, the THz image featured a reasonable cancer border and indicated that the tumour was composed of two tissue subtypes.
"We were expecting contrast between cancer and healthy parts but we were surprised to observe clear differences within cancer tissues," said Formanek. "Our goal is to work in close collaboration with a number of pathologists to build a complete database linking specific THz spectroscopic information with different type of tissues. The database could then be used in combination with automated analysis software to distinguish cancers from healthy tissues and even to identify different subtypes of cancers."
• Related articles in PMB
Terahertz imaging applied to cancer diagnosis
M-A Brun et al Phys. Med. Biol. 55 4615
Tissue characterization using terahertz pulsed imaging in reflection geometry
S Y Huang et al Phys. Med. Biol. 54 149
Assessing skin hydration status in haemodialysis patients using terahertz spectroscopy: a pilot/feasibility study
Filip Kadlec et al Phys. Med. Biol. 53 7063
The impact of hydration changes in fresh bio-tissue on THz spectroscopic measurements
G M Png et al Phys. Med. Biol. 53 3501