Deep learning‐based virtual noncalcium imaging in multiple myeloma using dual‐energy CT

Abstract

Background: Dual-energy CT with virtual noncalcium (VNCa) images allows the evaluation of focal intramedullary bone marrow involvement in patients with multiple myeloma. However, current commercial VNCa techniques suffer from excessive image noise and artifacts due to material decomposition used in synthesizing VNCa images. Objectives: In this work, we aim to improve VNCa image quality for the assessment of focal multiple myeloma, using an Artificial intelligence based Generalizable Algorithm for mulTi-Energy CT (AGATE) method. Materials and methods: AGATE method used a custom dual-task convolutional neural network (CNN) that concurrently carries out material classification and quantification. The material classification task provided an auxiliary regularization to the material quantification task. CNN parameters were optimized using custom loss functions that involved cross-entropy, physics-informed constraints, structural redundancy in spectral and material images, and texture information in spectral images. For training data, CT phantoms (diameters 30 to 45 cm) with tissue-mimicking inserts were scanned on a third generation dual-source CT system. Scans were performed at routine dose and half of the routine dose. Small image patches (i.e., 40 × 40 pixels) of tissue-mimicking inserts with known basis material densities were extracted for training samples. Numerically simulated insert materials with various shapes increased diversity of training samples. Generalizability of AGATE was evaluated using CT images from phantoms and patients. In phantoms, material decomposition accuracy was estimated using mean-absolute-percent-error (MAPE), using physical inserts that were not used during the training. Noise power spectrum (NPS) and modulation transfer function (MTF) were compared across phantom sizes and radiation dose levels. Five patients with multiple myeloma underwent dual-energy CT, with VNCa images generated using a commercial method and AGATE. Two fellowship-trained musculoskeletal radiologists reviewed the VNCa images (commercial and AGATE) side-by-side using a dual-monitor display, blinded to VNCa type, rating the image quality for focal multiple myeloma lesion visualization using a 5-level Likert comparison scale (−2 = worse visualization and diagnostic confidence, −1 = worse visualization but equivalent diagnostic confidence, 0 = equivalent visualization and diagnostic confidence, 1 = improved visualization but equivalent diagnostic confidence, 2 = improved visualization and diagnostic confidence). A post hoc assignment of comparison ratings was performed to rank AGATE images in comparison to commercial ones.

The evolution and current situation in the application of dual-energy computed tomography: a bibliometric study

“…In terms of image reconstruction, some studies have found that low-kiloelectron volt VMI reconstruction via deep learning can improve image quality in the evaluation of hypoenhanced hepatic metastasis and other liver diseases ( 35 - 37 ). Furthermore, deep learning can automatically segment brain tumors and surrounding healthy tissues to assist in the diagnosis of multiple myeloma ( 38 , 39 ). It can even predict head and neck lymph node metastasis of papillary thyroid carcinoma on DECT images ( 40 ).…”

Systematic Review on Learning-Based Spectral CT

“…Several deep learning architectures, previously designed to solve other image processing tasks, have been deployed for image-based decomposition. Most works are based on a supervised learning approach where a dataset of manually segmented basis material images are available: fully convolutional network [127], U-Net [128]- [133], Butterfly-Net [134], visual geometry group [132], [135], Incept-net [136], [137], generative adversarial network (GAN) [138], Dense-net [139]. These contributions differ on the type of architecture adopted and the complexity of the network which is measured by the number of trainable parameters.…”

“…They also differ in which inputs are used by the network, e.g., reconstructed multichannel CT images µ [133] or pre-decomposed CT images [131]. The network output is generally the decomposed CT images x m but it may also be other images, e.g., the elemental composition [132], quantities used for radiotherapy planning such as the image of the electron density [140] or the virtual non-calcium image [137].…”

Imaging of Multiple Myeloma: Present and Future

“…In terms of LDWBCT for the diagnosis of MM, artificial intelligence is being developed to generate tools that automatically perform segmentation and bone subtraction both in monoenergetic CT and DECT to recognize focal lesions, allowing for the quicker interpretation of results and increasing diagnostic accuracy [ 21 , 100 , 101 , 102 ], as well as creating convolutional neural networks that allow for a straight estimation of the distribution of materials, avoiding the need to perform the conventional material decomposition [ 103 ]. The development of algorithms to minimize noise and artifacts in VNCa images has yielded promising first results [ 104 ], and deep learning for photon-counting detector images has shown a greater performance in highlighting focal lesions [ 99 ].…”