What is medical image processing?
|
Balaji Ganeshan
|
Medical image processing is the branch of medical imaging associated with quantitative analysis and visualisation of medical images of numerous modalities such as Positron Emission Tomography (PET), Magnetic Resonance Imaging (MRI), Computed Tomography (CT), or microscopy to extract, enhance and display information that could be used by medical imaging professions (engineers, physicist and clinicians – radiologists) to diagnose, monitor and treat medical disorders.
What led you to analyse textures of radiological images?
Heterogeneity of the tumour microenvironment (e.g. tumour blood supply) is a well-recognised feature of malignancy that is associated with adverse tumour biology. A heterogeneous blood supply will also impact on treatment response due to poor delivery of chemotherapeutic agents to areas of low vascularity. Hence, a non-invasive imaging method for assessment of tumour heterogeneity could potentially provide a biomarker for prognosis and treatment response. Visual analysis of diagnostic images is largely based upon evaluating morphological information such as size and shape. Image perception and identifying relationships between perceived patterns and possible diagnosis heavily depend on radiologist’s knowledge, analytical skills, memory, intuition and diligence. However, the human visual system has difficulties in discriminating textural information such as coarseness and regularity that result from local spatial variations in image brightness. Furthermore, quantitative information from images is becoming increasingly important within radiological practice as it makes the process more objective. This however is not possible through visual analysis and therefore requires computer-based algorithmic processing. Texture analysis is a vital component of medical image analysis because it is difficult to classify human tissues via visual assessment based on shape or grey-level information. Also, improvements in texture analysis techniques would increase the extracted information enabling better quantification of differences in appearance inaccessible to the naked eye.
How was TexRAD born?
Most initial texture analysis work (prior to TexRAD) was focussed on tissue segmentation and identification of tissue as benign or malignant. There was very little work on assessing prognosis, disease-severity and treatment-response/prediction, which was more challenging and less developed. This was in fact the basis of my PhD research at the University of Sussex which started in October 2004 to develop a texture analysis algorithm for patient risk-stratification (‘personalised-medicine’), which could be used as an adjunct (confident decision-making) in routine clinical (radiological) practice. The texture analysis algorithm developed was novel in its approach within medical imaging and a patent application was made in 2007 to protect the invention (currently granted in a few jurisdictions). This generated some initial interest among few prestigious institutions in UK and Europe which led to us developing a research software prototype of the texture analysis. Further interest and scientific publications demonstrating its usefulness in cancer imaging and lack of a commercially available texture analysis software platform led to the spinning out of the company (TexRAD www.texrad.org was incorporated in February 2011) with a number of partnering companies, university and individuals:
- Imaging Equipment (Distributor of radiopharmaceuticals Nick Stevens – Managing Director)
- University of Sussex (Dr Ian Carter – University Director)
- w Dr Balaji Ganeshan (Scientific Director, Inventor, Senior Research Associate at the University College London)
- Cambridge Computed Imaging (medical software developing company – Mike Hayball, Technical Director)
- Miles Medical Pty (Prof Ken Miles – Consultant Radiologist & Nuclear Medicine Physician, Co-inventor, Professor of Medical Imaging, University College London)
The current academic base for TexRAD is at the Institute of Nuclear-Medicine, University College, London.
What does the TexRAD software do? How does it do it?
TexRAD is a novel measurement tool that enhances the ability of diagnostic imaging (CT, PET, MRI) to contribute to treatment decisions for patients with cancer and other diseases. To date, diagnostic imaging systems have generally enhanced fine detail with the aim of optimising anatomical resolution. However, experience from the automated identification of military targets has indicated that important discriminatory information is to be found within coarser variations in image brightness. If pronounced, these variations can be perceived as abnormalities of texture. TexRAD employs filters to specifically highlight these coarser features (radii 2 -12mm) and uses histogram analysis to quantify the filtered images.
How is this information used for risk assessment of a cancerous tissue?
Clinical research applying TexRAD to a range of tumours (lung, oesophageal, colorectal, breast, prostate, renal cell cancer) has identified biological correlates of known prognostic significance and shown the ability of cross-validated threshold texture values to stratify patients by prognosis and/or treatment response.
What are the other ways in which this software helps oncologists/researchers?
Modelling studies have demonstrated that the use of TexRAD to analyse medical images for quantifying tumour heterogeneity acquired in routine clinical practice can potentially impact clinical decision and assist the clinicians (e.g. oncologists), making, for example, suitability for cancer chemotherapy (use the TexRAD information as an adjunct/additional prognostic factors in Adjuvant! Online), improving the ability to accommodate patient preferences and to save costs of inappropriate treatment that might have been selected using existing methods.
Why is information on tumour heterogeneity crucial to oncologists today? This software is particularly generating a lot of interest in doctors working with thoracic malignancies, renal cell carcinoma, haepatocellular carcinoma. Why?
These malignancies are some of the most common cancers worldwide, and in India. Their prognosis is also poorer. With early risk-stratification and optimised treatment, the overall outcome for patients with these cancers can be potentially improved. Imaging generally forms part of the first diagnostic test in detecting these cancers. Hence there will be great value if novel imaging biomarkers can be developed for early risk-stratification.
Research using TexRAD on CT has shown the ability to sub-select metastatic renal cell cancer patients who will respond well to a specific type of targeted therapy (anti-angiogenic drugs which are not only expensive but toxic) from those who will not; for whom another form of treatment may be beneficial.
Research using TexRAD on CT has shown the ability to identify poor prognostic (reduced-survival) lung cancer patients (at the time of staging) and haepatocellular carcinoma patients from good prognostic cases, potentially assisting the clinician to optimise treatment strategies for better patient outcome.
What has been the customer experience with TexRAD? Are any hospitals in India using this software?
We have been working with prestigious clinical and research institutions around the world and the interest and feedback has been encouraging. The interest in TexRAD has been gaining a lot of momentum as a novel research tool (to enhance research output and establish novel clinical applications) leading to the development of a potentially useful clinical tool. This is evident from the increasing number of high-impact research papers and conference publications from the TexRAD user community, an indication of the acceptance of TexRAD within the scientific community. (http://www.texrad.org/index.php?option=com_content&view=article&id=4&Itemid=5)
Additionally, the very recent research work undertaken by Tata Memorial Hospital has recently shown the potential application in cervical cancer prognosis and response assessment.
Anything else you would like to add.
This high level of performance verification undertaken for TexRAD is unusual, if not unique, amongst imaging biomarkers. TexRAD does not require specialised imaging protocols to be added to existing imaging and the barriers to uptake are therefore likely to be low. TexRAD has comparable or superior prognostic performance and lower cost than serum or pathological biomarkers. TexRAD is therefore well placed to fulfil the need for readily available prognostic biomarkers to underpin stratified medicine.