Can portable tomosynthesis improve the diagnostic value of bedside chest X-ray in the intensive care unit? A proof of concept study

AI-based computer-aided diagnostic system of chest digital tomography synthesis: Demonstrating comparative advantage with X-ray-based AI systems

BACKGROUND

Compared with chest X-ray (CXR) imaging, which is a single image projected from the front of the patient, chest digital tomosynthesis (CDTS) imaging can be more advantageous for lung lesion detection because it acquires multiple images projected from multiple angles of the patient. Various clinical comparative analysis and verification studies have been reported to demonstrate this, but there is no artificial intelligence (AI)-based comparative analysis studies. Existing AI-based computer-aided detection (CAD) systems for lung lesion diagnosis have been developed mainly based on CXR images; however, CAD-based on CDTS, which uses multi-angle images of patients in various directions, has not been proposed and verified for its usefulness compared to CXR-based counterparts.

BACKGROUND AND OBJECTIVE

This study develops and tests a CDTS-based AI CAD system to detect lung lesions to demonstrate performance improvements compared to CXR-based AI CAD.

METHODS

We used multiple (e.g., five) projection images as input for the CDTS-based AI model and a single-projection image as input for the CXR-based AI model to compare and evaluate the performance between models. Multiple/single projection input images were obtained by virtual projection on the three-dimensional (3D) stack of computed tomography (CT) slices of each patient's lungs from which the bed area was removed. These multiple images result from shooting from the front and left and right 30/60∘. The projected image captured from the front was used as the input for the CXR-based AI model. The CDTS-based AI model used all five projected images. The proposed CDTS-based AI model consisted of five AI models that received images in each of the five directions, and obtained the final prediction result through an ensemble of five models. Each model used WideResNet-50. To train and evaluate CXR- and CDTS-based AI models, 500 healthy data, 206 tuberculosis data, and 242 pneumonia data were used, and three three-fold cross-validation was applied.

RESULTS

The proposed CDTS-based AI CAD system yielded sensitivities of 0.782 and 0.785 and accuracies of 0.895 and 0.837 for the (binary classification) performance of detecting tuberculosis and pneumonia, respectively, against normal subjects. These results show higher performance than the sensitivity of 0.728 and 0.698 and accuracies of 0.874 and 0.826 for detecting tuberculosis and pneumonia through the CXR-based AI CAD, which only uses a single projection image in the frontal direction. We found that CDTS-based AI CAD improved the sensitivity of tuberculosis and pneumonia by 5.4% and 8.7% respectively, compared to CXR-based AI CAD without loss of accuracy.

CONCLUSIONS

This study comparatively proves that CDTS-based AI CAD technology can improve performance more than CXR. These results suggest that we can enhance the clinical application of CDTS. Our code is available at https://github.com/kskim-phd/CDTS-CAD-P.

Numerical methods for CT reconstruction with unknown geometry parameters

Computed tomography (CT) techniques are well known for their ability to produce high-quality images needed for medical diagnostic purposes. Unfortunately, standard CT machines are extremely large, heavy, require careful and regular calibration, and are expensive, which can limit their availability in point-of-care situations. An alternative approach is to use portable machines, but parameters related to the geometry of these devices (e.g., distance between source and detector, orientation of source to detector) cannot always be precisely calibrated, and these parameters may change slightly when the machine is adjusted during the image acquisition process. In this work, we describe the non-linear inverse problem that models this situation, and discuss algorithms that can jointly estimate the geometry parameters and compute a reconstructed image. In particular, we propose a hybrid machine learning and block coordinate descent (ML-BCD) approach that uses an ML model to calibrate geometry parameters, and uses BCD to refine the predicted parameters and reconstruct the imaged object simultaneously. We show using numerical experiments that our new method can efficiently improve the accuracy of both the image and geometry parameters.

Digital tomography in the diagnosis of a posterior pneumothorax in the intensive care unit

Portable chest radiography is a valuable tool in the intensive care unit. However, the supine position causes superposition of anatomical structures resulting in less reliable detection of certain abnormalities. Recently, a portable digital tomosynthesis (pDTS) prototype with a modified motorized X-ray device was developed. We aimed to compare the diagnostic value of pDTS to standard bedside chest radiography in the diagnosis of a posterior pneumothorax. A modified motorized X-ray device was developed to perform 15 radiographic projections while translating the X-ray tube 25 cm (10 cm ramp up and 15 cm during X-ray exposure) with a total radiation dose of 0.54 mSv. This new technique of pDTS was performed in addition to standard bedside chest X-ray in a patient with a confirmed posterior hydropneumothorax. The images were compared with the standard bedside chest X-ray and computed tomography (CT) images by two experienced radiologists. The posterior hydropneumothorax previously identified with CT was visible on tomosynthesis images but not with standard bedside imaging. Combining the digital tomosynthesis technique with the portable X-ray machine could increase the diagnostic value of bedside chest radiography for the diagnosis of posterior pneumothoraces while avoiding intrahospital transport and limiting radiation exposure compared to CT.

3D chest tomosynthesis using a stationary flat panel source array and a stationary detector: a Monte Carlo proof of concept

3D imaging modalities such as computed tomography and digital tomosynthesis typically scan the patient from different angles with a lengthy mechanical movement of a single x-ray tube. Therefore, millions of 3D scans per year require expensive mechanisms to support a heavy x-ray source and have to compensate for machine vibrations and patient movements. However, recent developments in cold-cathode field emission technology allow the creation of compact, stationary arrays of emitters. Adaptix Ltd has developed a novel, low-cost, square array of such emitters and demonstrated 3D digital tomosynthesis of human extremities and small animals. The use of cold-cathode field emitters also makes the system compact and lightweight. This paper presents Monte Carlo simulations of a concept upgrade of the Adaptix system from the current 60 kVp to 90 kVp and 120 kVp which are better suited for chest imaging. Between 90 kVp and 120 kVp, 3D image quality appears insensitive to voltage and at 90 kVp the photon yield is reduced by 40%-50% while effective dose declines by 14%. A square array of emitters can adequately illuminate a subject for tomosynthesis from a shorter source-to-image distance, thereby reducing the required input power, and offsetting the 28%-50% more input power that is required for operation at 90 kVp. This modelling suggests that lightweight, stationary cold-cathode x-ray source arrays could be used for chest tomosynthesis at a lower voltage, with less dose and without sacrificing image quality. This will reduce weight, size and cost, enabling 3D imaging to be brought to the bedside.