Lung cancer is considered the most common cancer globally, and has been for the past decades. Its prevalence is found especially high in developed/Westernized countries, the highest being Eastern Europe and Eastern Asia. It is unfortunately a very deadly disease, with numbers showing less than 6 out of 10 patients surviving more than 5 years even if the patient is diagnosed in the earliest stage of the disease.
Being very aggressive, lung cancer develops into a metastatic disease quickly and is also hard to treat. It is therefore of great importance that once diagnosed, lung cancer is completely characterized. Only when the extent of the disease is accurately known can the disease be treated accurately. This means not only a primary lesion suspected of lung cancer needs to be evaluated but also all lymph nodes surrounding the lungs need to be assessed. To do so, staging is generally performed by means of a flexible endoscope with an ultrasound transducer at the distal tip; EBUS-TBNA (endobronchial ultrasound – transbronchial needle aspiration) and EUS-FNA (endosophageal ultrasound – fine needle aspiration). Under guidance of ultrasound, the physician will assess all lymph nodes inside the mediastinum and around the lungs as adjacent to the esophagus and main airways. If ultrasound imaging suspicious nodes are then found, they are aspirated by a hollow needle as inserted through the bronchoscope.
The needle aspiration of lymph nodes by EBUS-TBNA and EUS-FNA has shown to be a relatively accurate means of staging lung cancer. However, it is sometimes troubled by findings that multiple lymph nodes inside a single anatomical region are visible. The physician then needs to decide on which nodes to aspirate. . Additionally, while quite accurate, it is not seldomly found post-procedurally that not enough cells have been collected for enabling a diagnosis of the aspired lymph node.
The readily available ultrasound imaging information collected during the EBUS-TBNA and EUS-FNA procedures might herein be excellently used to further help determine risk of malignancy. Therefore, our research focuses on evaluating how using ultrasound information might improve lung cancer staging. Main topic herein include the (1) implementation of ultrasound strain elastography for determining if relative elasticity of lymph nodes enables a malignancy prediction, (2) using traditional b-mode imaging features, (3) quantitative CAD systems and (4) integration of ultrasound information with other available imaging information for predicting chance of lymph node malignancy (i.e. PET-CT and/or CT).
Ultrasound strain elastography is a technique which uses subsequently acquired RF signals to correlate pre- and post- compression deformation into a strain map. In other words, it approximates the relative stiffness of tissue that is found in the axial direction of the US probe through correlation of an RF/US image before- and after mechanical compression. The strain map, shows relative in-image stiffness of tissue. Since it is semi-quantitative, we have developed and tested a clinical measurement protocol for being able to acquire these measurements. We have furthermore initiated and performed an international multi-center trial using commercially available equipment for testing if it helps predict lymph node malignancy in a semi-quantitative way, with positive results.
Aside of strain elastography, B-mode features are routinely observed by the physician. These have historically been proposed as good predictors of malignancy. By designing an international multi-center trial, we have prospectively assessed 525 lymph nodes for these features. We herein found that the clinical applicability in a subjective setting is limited. While individual features did show of value, an observer scoring study we further initiated found that the variability in observers might be to blame that the compounding of features does not result in a highly accurate prediction of malignancy. Future work will now focus on designing automated computer algorithms that can objectively and reproducibly score these features. and, will further integrate ultrasound features not easily quantified by the physician.
An EBUS-TBNA and EUS-FNA procedure will only be performed once a PET-CT and/or CT has shown a lung cancer suspicion along with nodal involvement. The information that is harbored within the combination of modalities along with initial patient information might uniquely provide a means for more accurate prediction of disease state and possibly also characteristics such as disease progression. With retrospective data on all these available, we are now assessing if artificial intelligence algorithms can be developed for helping us decide on disease staging and risk stratification of disease.
- Commercially funded research