The power of feedback.

In recent years several review articles and books have been published on the use of porphyrin-based compounds in photodynamic therapy (PDT). This critical review is focused on (i) the basic concept of PDT, (ii) advantages of long-wavelength absorbing photosensitizers (PS), (iii) a brief discussion on recent advances in developing PDT agents, and (iv) the various synthetic strategies designed at the Roswell Park Cancer Institute, Buffalo, for developing highly effective long-wavelength PDT agents and their utility in constructing the conjugates with tumor-imaging and therapeutic potential (Theranostics). The clinical status of certain selected PDT agents is also summarized (205 references).

The role of porphyrin chemistry in tumor imaging and photodynamic therapy

Deformable image registration is a fundamental task in medical image processing. Among its most important applications, one may cite: 1) multi-modality fusion, where information acquired by different imaging devices or protocols is fused to facilitate diagnosis and treatment planning; 2) longitudinal studies, where temporal structural or anatomical changes are investigated; and 3) population modeling and statistical atlases used to study normal anatomical variability. In this paper, we attempt to give an overview of deformable registration methods, putting emphasis on the most recent advances in the domain. Additional emphasis has been given to techniques applied to medical images. In order to study image registration methods in depth, their main components are identified and studied independently. The most recent techniques are presented in a systematic fashion. The contribution of this paper is to provide an extensive account of registration techniques in a systematic manner.

/pdf/deformable-medical-image-registration-a-survey-2uee59wp1z.pdf

Deformable Medical Image Registration: A Survey

What has happened in machine learning lately, and what does it mean for the future of medical image analysis? Machine learning has witnessed a tremendous amount of attention over the last few years. The current boom started around 2009 when so-called deep artificial neural networks began outperforming other established models on a number of important benchmarks. Deep neural networks are now the state-of-the-art machine learning models across a variety of areas, from image analysis to natural language processing, and widely deployed in academia and industry. These developments have a huge potential for medical imaging technology, medical data analysis, medical diagnostics and healthcare in general, slowly being realized. We provide a short overview of recent advances and some associated challenges in machine learning applied to medical image processing and image analysis. As this has become a very broad and fast expanding field we will not survey the entire landscape of applications, but put particular focus on deep learning in MRI. 
Our aim is threefold: (i) give a brief introduction to deep learning with pointers to core references; (ii) indicate how deep learning has been applied to the entire MRI processing chain, from acquisition to image retrieval, from segmentation to disease prediction; (iii) provide a starting point for people interested in experimenting and perhaps contributing to the field of machine learning for medical imaging by pointing out good educational resources, state-of-the-art open-source code, and interesting sources of data and problems related medical imaging.

An overview of deep learning in medical imaging focusing on MRI

Purpose
Interests have been rapidly growing in the field of radiotherapy to replace CT with magnetic resonance imaging (MRI), due to superior soft tissue contrast offered by MRI and the desire to reduce unnecessary radiation dose. MR-only radiotherapy also simplifies clinical workflow and avoids uncertainties in aligning MR with CT. Methods, however, are needed to derive CT-equivalent representations, often known as synthetic CT (sCT), from patient MR images for dose calculation and DRR-based patient positioning. Synthetic CT estimation is also important for PET attenuation correction in hybrid PET-MR systems. We propose in this work a novel deep convolutional neural network (DCNN) method for sCT generation and evaluate its performance on a set of brain tumor patient images.

Methods
The proposed method builds upon recent developments of deep learning and convolutional neural networks in the computer vision literature. The proposed DCNN model has 27 convolutional layers interleaved with pooling and unpooling layers and 35 million free parameters, which can be trained to learn a direct end-to-end mapping from MR images to their corresponding CTs. Training such a large model on our limited data is made possible through the principle of transfer learning and by initializing model weights from a pretrained model. Eighteen brain tumor patients with both CT and T1-weighted MR images are used as experimental data and a sixfold cross-validation study is performed. Each sCT generated is compared against the real CT image of the same patient on a voxel-by-voxel basis. Comparison is also made with respect to an atlas-based approach that involves deformable atlas registration and patch-based atlas fusion.

Results
The proposed DCNN method produced a mean absolute error (MAE) below 85 HU for 13 of the 18 test subjects. The overall average MAE was 84.8 ± 17.3 HU for all subjects, which was found to be significantly better than the average MAE of 94.5 ± 17.8 HU for the atlas-based method. The DCNN method also provided significantly better accuracy when being evaluated using two other metrics: the mean squared error (188.6 ± 33.7 versus 198.3 ± 33.0) and the Pearson correlation coefficient(0.906 ± 0.03 versus 0.896 ± 0.03). Although training a DCNN model can be slow, training only need be done once. Applying a trained model to generate a complete sCT volume for each new patient MR image only took 9 s, which was much faster than the atlas-based approach.

Conclusions
A DCNN model method was developed, and shown to be able to produce highly accurate sCT estimations from conventional, single-sequence MR images in near real time. Quantitative results also showed that the proposed method competed favorably with an atlas-based method, in terms of both accuracy and computation speed at test time. Further validation on dose computation accuracy and on a larger patient cohort is warranted. Extensions of the method are also possible to further improve accuracy or to handle multi-sequence MR images.

MR-based synthetic CT generation using a deep convolutional neural network method

Overview of image-guided radiation therapy.

On-board CBCT images are used to generate patient geometric models to assist patient setup. The image data can also, potentially, be used for dose reconstruction in combination with the fluence maps from treatment plan. Here we evaluate the achievable accuracy in using a kV CBCT for dose calculation. Relative electron density as a function of HU was obtained for both planning CT (pCT) and CBCT using a Catphan-600 calibration phantom. The CBCT calibration stability was monitored weekly for 8 consecutive weeks. A clinical treatment planning system was employed for pCT- and CBCT-based dose calculations and subsequent comparisons. Phantom and patient studies were carried out. In the former study, both Catphan-600 and pelvic phantoms were employed to evaluate the dosimetric performance of the full-fan and half-fan scanning modes. To evaluate the dosimetric influence of motion artefacts commonly seen in CBCT images, the Catphan-600 phantom was scanned with and without cyclic motion using the pCT and CBCT scanners. The doses computed based on the four sets of CT images (pCT and CBCT with/without motion) were compared quantitatively. The patient studies included a lung case and three prostate cases. The lung case was employed to further assess the adverse effect of intra-scan organ motion. Unlike the phantom study, the pCT of a patient is generally acquired at the time of simulation and the anatomy may be different from that of CBCT acquired at the time of treatment delivery because of organ deformation. To tackle the problem, we introduced a set of modified CBCT images (mCBCT) for each patient, which possesses the geometric information of the CBCT but the electronic density distribution mapped from the pCT with the help of a BSpline deformable image registration software. In the patient study, the dose computed with the mCBCT was used as a surrogate of the 'ground truth'. We found that the CBCT electron density calibration curve differs moderately from that of pCT. No significant fluctuation was observed in the calibration over the period of 8 weeks. For the static phantom, the doses computed based on pCT and CBCT agreed to within 1%. A notable difference in CBCT- and pCT-based dose distributions was found for the motion phantom due to the motion artefacts which appeared in the CBCT images (the maximum discrepancy was found to be approximately 3.0% in the high dose region). The motion artefacts-induced dosimetric inaccuracy was also observed in the lung patient study. For the prostate cases, the mCBCT- and CBCT-based dose calculations yielded very close results (<2%). Coupled with the phantom data, it is concluded that the CBCT can be employed directly for dose calculation for a disease site such as the prostate, where there is little motion artefact. In the prostate case study, we also noted a large discrepancy between the original treatment plan and the CBCT (or mCBCT)-based calculation, suggesting the importance of inter-fractional organ movement and the need for adaptive therapy to compensate for the anatomical changes in the future.

http://iopscience.iop.org/article/10.1088/0031-9155/52/3/011/pdf

Evaluation of on-board kV cone beam CT (CBCT)-based dose calculation*

Positron emission tonography (PET) is useful in diagnosis and radiation treatment planning for a variety of cancers. For patients with cancers in thoracic or upper abdominal region, the respiratory motion produces large distortions in the tumor shape and size, affecting the accuracy in both diagnosis and treatment. Four-dimensional (4D) (gated) PET aims to reduce the motion artifacts and to provide accurate measurement of the tumor volume and the tracer concentration. A major issue in 4D PET is the lack of statistics. Since the collected photons are divided into several frames in the 4D PET scan, the quality of each reconstructed frame degrades as the number of frames increases. The increased noise in each frame heavily degrades the quantitative accuracy of the PET imaging. In this work, we propose a method to enhance the performance of 4D PET by developing a new technique of 4D PET reconstruction with incorporation of an organ motion model derived from 4D-CT images. The method is based on the well-known maximum-likelihood expectation-maximization (ML-EM) algorithm. During the processes of forward- and backward-projection in the ML-EM iterations, all projection data acquired at different phases are combined together to update the emission map with the aid of deformable model, the statistics is therefore greatly improved. The proposed algorithm was first evaluated with computer simulations using a mathematical dynamic phantom. Experiment with a moving physical phantom was then carried out to demonstrate the accuracy of the proposed method and the increase of signal-to-noise ratio over three-dimensional PET. Finally, the 4D PET reconstruction was applied to a patient case.

Model-based image reconstruction for four-dimensional PET

On-board imager (OBI) based cone-beam computed tomography (CBCT) has become available in radiotherapy clinics to accurately identify the target in the treatment position. However, due to the relatively slow gantry rotation (typically about 60 s for a full 360° scan) in acquiring the CBCT projection data, the patient's respiratory motion causes serious problems such as blurring, doubling, streaking and distortion in the reconstructed images, which heavily degrade the image quality and the target localization. In this work, we present a motion compensation method for slow-rotating CBCT scans by incorporating into image reconstruction a patient-specific motion model, which is derived from previously obtained four-dimensional (4D) treatment planning CT images of the same patient via deformable registration. The registration of the 4D CT phases results in transformations representing a temporal sequence of three-dimensional (3D) deformation fields, or in other words, a 4D model of organ motion. The algorithm was developed heuristically in two-dimensional (2D) parallel-beam geometry and extended to 3D cone-beam geometry. By simulations with digital phantoms capable of translational motion and other complex motion, we demonstrated that the algorithm can reduce the motion artefacts locally, and restore the tumour size and shape, which may thereby improve the accuracy of target localization and patient positioning when CBCT is used as the treatment guidance.

http://iopscience.iop.org/article/10.1088/0031-9155/51/2/005/pdf

Motion correction for improved target localization with on-board cone-beam computed tomography

The gastrin-releasing peptide receptor (GRPR) is found to be overexpressed in a variety of human tumors. The aim of this study was to develop 18F-labeled bombesin analogs for PET of GRPR expression in prostate cancer xenograft models. Methods: [Lys3]Bombesin ([Lys3]BBN) and aminocaproic acid-bombesin(7–14) (Aca-BBN(7–14)) were labeled with 18F by coupling the Lys3 amino group and Aca amino group, respectively, with N-succinimidyl-4-18F-fluorobenzoate (18F-SFB) under slightly basic condition (pH 8.5). Receptor-binding affinity of FB-[Lys3]BBN and FB-Aca-BBN(7–14) was tested in PC-3 human prostate carcinoma cells. Internalization and efflux of both radiotracers were also evaluated. Tumor-targeting efficacy and in vivo kinetics of both radiotracers were examined in male athymic nude mice bearing subcutaneous PC-3 tumors by means of biodistribution and dynamic microPET imaging studies. 18F-FB-[Lys3]BBN was also tested for orthotopic PC-3 tumor delineation. Metabolic stability of 18F-FB-[Lys3]BBN was determined in mouse blood, urine, liver, kidney, and tumor homogenates at 1 h after injection. Results: The typical decay-corrected radiochemical yield was about 30%–40% for both tracers, with a total reaction time of 150 ± 20 min starting from 18F−. 18F-FB-[Lys3]BBN had moderate stability in the blood and PC-3 tumor, whereas it was degraded rapidly in the liver, kidneys, and urine. Both radiotracers exhibited rapid blood clearance. 18F-FB-[Lys3]BBN had predominant renal excretion. 18F-FB-Aca-BBN(7–14) exhibited both hepatobiliary and renal clearance. Dynamic microPET imaging studies revealed that the PC-3 tumor uptake of 18F-FB-[Lys3]BBN in PC-3 tumor was much higher than that of 18F-FB-Aca-BBN(7–14) at all time points examined (P

https://jnm.snmjournals.org/content/47/3/492.full.pdf

Eduard Schreibmann

Papers

Overview of image-guided radiation therapy.

Evaluation of on-board kV cone beam CT (CBCT)-based dose calculation*

Model-based image reconstruction for four-dimensional PET

Motion correction for improved target localization with on-board cone-beam computed tomography

18F-Labeled Bombesin Analogs for Targeting GRP Receptor-Expressing Prostate Cancer