Original Computed Tomography Scan Research Articles

Breaking boundaries in radiology: redefining AI diagnostics via raw data ahead of reconstruction

Objective. In the realm of utilizing artificial intelligence (AI) for medical image analysis, the paradigm of ‘signal-image-knowledge’ has remained unchanged. However, the process of ‘signal to image’ inevitably introduces information distortion, ultimately leading to irrecoverable biases in the ‘image to knowledge’ process. Our goal is to skip reconstruction and build a diagnostic model directly from the raw data (signal). Approach. This study focuses on computed tomography (CT) and its raw data (sinogram) as the research subjects. We simulate the real-world process of ‘human-signal-image’ using the workflow ‘CT-simulated data- reconstructed CT,’ and we develop a novel AI predictive model directly targeting raw data (RCTM). This model comprises orientation, spatial, and global analysis modules, embodying the fusion of local to global information extraction from raw data. We selected 1994 patients with retrospective cases of solid lung nodules and modeled different types of data. Main results. We employed predefined radiomic features to assess the diagnostic feature differences caused by reconstruction. The results indicated that approximately 14% of the features had Spearman correlation coefficients below 0.8. These findings suggest that despite the increasing maturity of CT reconstruction algorithms, they still introduce perturbations to diagnostic features. Moreover, our proposed RCTM achieved an area under the curve (AUC) of 0.863 in the diagnosis task, showcasing a comprehensive superiority over models constructed from secondary reconstructed CTs (0.840, 0.822, and 0.825). Additionally, the performance of RCTM closely resembled that of models constructed from original CT scans (0.868, 0.878, and 0.866). Significance. The diagnostic and therapeutic approach directly based on CT raw data can enhance the precision of AI models and the concept of ‘signal-to-image’ can be extended to other types of imaging. AI diagnostic models tailored to raw data offer the potential to disrupt the traditional paradigm of ‘signal-image-knowledge’, opening up new avenues for more accurate medical diagnostics.

Pancreas segmentation using a dual-input v-mesh network.

Accurate segmentation of the pancreas from abdomen scans is crucial for the diagnosis and treatment of pancreatic diseases. However, the pancreas is a small, soft and elastic abdominal organ with high anatomical variability and has a low tissue contrast in computed tomography (CT) scans, which makes segmentation tasks challenging. To address this challenge, we propose a dual-input v-mesh fully convolutional network (FCN) to segment the pancreas in abdominal CT images. Specifically, dual inputs, i.e., original CT scans and images processed by a contrast-specific graph-based visual saliency (GBVS) algorithm, are simultaneously sent to the network to improve the contrast of the pancreas and other soft tissues. To further enhance the ability to learn context information and extract distinct features, a v-mesh FCN with an attention mechanism is initially utilized. In addition, we propose a spatial transformation and fusion (SF) module to better capture the geometric information of the pancreas and facilitate feature map fusion. We compare the performance of our method with several baseline and state-of-the-art methods on the publicly available NIH dataset. The comparison results show that our proposed dual-input v-mesh FCN model outperforms previous methods in terms of the Dice similarity coefficient (DSC), positive predictive value (PPV), sensitivity (SEN), average surface distance (ASD) and Hausdorff distance (HD). Moreover, ablation studies show that our proposed modules/structures are critical for effective pancreas segmentation.

Dosimetric evaluation of image-guided radiation therapy for prostate cancer

The aim of this study was to investigate the dosimetric accuracy of imaged-guided radiation therapy for prostate patients using the in-room computed tomography (CT) target localization technique. A Siemens CT-on-rails system was used for patient setup and target localization for intensity-modulated radiation therapy (IMRT) of prostate cancer. Fifteen previously treated prostate patients were included in this retrospective study. CT-on-Rails scans were performed before and after the IMRT treatment under local IRB approval. A total of 15 original simulation CT scans and 98 post-treatment CT scans were contoured by the same oncologist to delineate the prostate target, bladder, and rectum. IMRT plans were generated on the original simulation CTs and the same MUs and leaf sequences were used to compute the dose distributions using post-treatment CTs. Varian Velocity deformable registration was used for the summation of the fractional doses and the cumulative doses were compared with the planned doses. For the 15 patients investigated, the mean isocenter shift was up to 4.0 mm in the left-right direction, 5.9 mm in the anterior-posterior direction and 5.6 mm in the superior-inferior direction due to interfractional organ motion. The mean rectal volume varied from 0.6 to 1.73 times and the mean bladder volume varied from 0.59 to 3.65 times between simulation and the end of treatment. The prescription dose to 95% of the PTV, Dp, was set to 76 Gy for all treatment plans. The dose to 95% of the clinical treatment volume (CTV), D95, was 74.0 to 77.6 Gy and the minimum CTV dose, Dmin, was 61.0 to 71.6 Gy, respectively, in the cumulative dose distributions. Detailed analyses showed that 7.1% of the treatment fractions had cold spots (< 85% of Dp) in the peripheral CTV, leading to Dmin < 64 Gy in the cumulative dose distributions for 4 patients. The rectal dose-volume constraints were violated in 35.7% of the treatment fractions while the bladder dose was much improved in 82.7% of the treatment fractions. The current IGRT procedure for patient setup and target localization using rigid-body registration based on contour/anatomy matching is effective for population-based PTV margins. For a small group of patients, specific PTV margins and/or real-time target monitoring/tracking will be necessary due to significant prostate deformation/rotation caused by inter- and intrafractional bladder and rectal volume variation.

Technical Note: Synthesizing of lung tumors in computed tomography images

When investigating new radiation therapy techniques in the treatment planning stage, it can be extremely time consuming to locate multiple patient scans that match the desired characteristics for the treatment. With the help of machine learning, we propose to bypass the difficulty in finding patient computed tomography (CT) scans that match the treatment requirements. Furthermore, we aim to provide the developed method as a tool that is easily accessible to interested researchers. We propose a generative adversarial network (GAN) to edit individual volumes of interest (VOIs) in pre-existing CT scans, translating features of the healthy VOIs into features of cancerous volumes. Training and testing was done using VOIs from a dataset of 460 diagnostic and lung cancer screening CT scans. Agreement between real tumors and those produced by the editor was tested by comparing the distributions of several histogram parameters and second-order statistics as well as using qualitative analysis. After training, the network was successfully able to map healthy CT segments to realistic looking cancerous volumes. Based on visual inspection, tumors produced by the editor were found to be both realistic and visually consistent with the surrounding anatomy when placed back into the original CT scan. Furthermore, the network was found to be able to extrapolate well beyond the upper size limit of the training set. Lastly, a graphical user interface (GUI) was developed to easily interact with the resulting network. The trained network and associated GUI can serve as a tool to develop an abundance of lung cancer patient data to be used in treatment planning. In addition, this method can be extended to a variety of cancer types if given an appropriate baseline dataset. The GUI and instructions on how to utilize the tool have been made publicly available at https://github.com/teaghan/CT_Editor.

Deep Learning for Lung Cancer Nodules Detection and Classification in CT Scans

Detecting malignant lung nodules from computed tomography (CT) scans is a hard and time-consuming task for radiologists. To alleviate this burden, computer-aided diagnosis (CAD) systems have been proposed. In recent years, deep learning approaches have shown impressive results outperforming classical methods in various fields. Nowadays, researchers are trying different deep learning techniques to increase the performance of CAD systems in lung cancer screening with computed tomography. In this work, we review recent state-of-the-art deep learning algorithms and architectures proposed as CAD systems for lung cancer detection. They are divided into two categories—(1) Nodule detection systems, which from the original CT scan detect candidate nodules; and (2) False positive reduction systems, which from a set of given candidate nodules classify them into benign or malignant tumors. The main characteristics of the different techniques are presented, and their performance is analyzed. The CT lung datasets available for research are also introduced. Comparison between the different techniques is presented and discussed.

Deep learning-based image quality improvement for low-dose computed tomography simulation in radiation therapy.

Low-dose computed tomography (CT) is desirable for treatment planning and simulation in radiation therapy. Multiple rescanning and replanning during the treatment course with a smaller amount of dose than a single conventional full-dose CT simulation is a crucial step in adaptive radiation therapy. We developed a machine learning-based method to improve image quality of low-dose CT for radiation therapy treatment simulation. We used a residual block concept and a self-attention strategy with a cycle-consistent adversarial network framework. A fully convolution neural network with residual blocks and attention gates (AGs) was used in the generator to enable end-to-end transformation. We have collected CT images from 30 patients treated with frameless brain stereotactic radiosurgery (SRS) for this study. These full-dose images were used to generate projection data, which were then added with noise to simulate the low-mAs scanning scenario. Low-dose CT images were reconstructed from this noise-contaminated projection data and were fed into our network along with the original full-dose CT images for training. The performance of our network was evaluated by quantitatively comparing the high-quality CT images generated by our method with the original full-dose images. When mAs is reduced to 0.5% of the original CT scan, the mean square error of the CT images obtained by our method is , with respect to the original full-dose images. The proposed method successfully improved the noise, contract-to-noise ratio, and nonuniformity level to be close to those of full-dose CT images and outperforms a state-of-the-art iterative reconstruction method. Dosimetric studies show that the average differences of dose-volume histogram metrics are ( ). These quantitative results strongly indicate that the denoised low-dose CT images using our method maintains image accuracy and quality and are accurate enough for dose calculation in current CT simulation of brain SRS treatment. We also demonstrate the great potential for low-dose CT in the process of simulation and treatment planning.

Metachronous or synchronous primary lung cancer in the era of computed tomography surveillance

ObjectivesThe purpose of this study was to determine the frequency, characteristics, and survival of second primary lung cancer initially identified as an indeterminate lesion on the original computed tomography scan and then diagnosed during the surveillance period in a prospective study. MethodsA prospective database of 271 patients enrolled in a surveillance study was updated. Indeterminate lesions present on the original computed tomography at the time of initial primary lung cancer diagnosis that subsequently grew and were diagnosed as cancer were termed “synchronous primary lung cancer.” Lesions that were not present on the original computed tomography scan and subsequently diagnosed on surveillance were termed “metachronous primary lung cancer.” ResultsThirty patients (11.1%) developed 37 second primary lung cancers over a median surveillance period of 84.7 (range, 15.9-147.6) months. Of these, 15 of 37 (40.5%) were identified as synchronous primary lung cancer, and 22 of 37 (59.5%) were identified as metachronous primary lung cancer. At first identification, ground-glass lesions were identified in 9 of 15 (60%) synchronous primary lung cancers compared with only 5 of 22 (22.7%) of metachronous primary lung cancers (P = .034). Compared with metachronous primary lung cancer, from first identification to diagnosis, synchronous primary lung cancer developed over a longer interval (33.6 vs 7.2 months, P = .001) and had a slower growth rate (0.17 vs 0.45 mm/month, P = .027). The 5-year overall survival from second lung cancer was 73.0%. No significant differences were observed between the synchronous primary lung cancer and metachronous primary lung cancer cohorts in overall survival from initial primary lung cancer (P = .583) or from second lung cancer (P = .966). ConclusionsComputed tomography surveillance identifies 2 types of curable second lung cancers leading to excellent overall survival.

Assessment of electron density effects on dose calculation and optimisation accuracy for nasopharynx, for MRI only treatment planning.

Computed tomography (CT) is the gold standard for radiotherapy simulation and treatment planning, providing spatial accuracy, bony anatomy definition and electron density information for dose calculations. Magnetic resonance imaging (MRI) has been introduced in radiotherapy to improve visualisation of anatomy for accurate target definition and contouring, however lacks electron density information required for dose calculations, with various methods used to overcome this. The aim of this work is to assess the impact on dose calculation accuracy and optimisation results of different approaches to determine electron density, as could be used in MRI only treatment planning for nasopharyngeal datasets with VMAT treatment plans. Volumetric modulated arc therapy (VMAT) plans were created for 10 retrospective head and neck (H&N) nasopharyngeal patients. The VMAT plans were generated on the gold standard dataset, the original CT scan. Data sets with no density correction (water equivalent) and two different sets of bulk density correction for bone/air/tissue applied separately were generated for these patients and the VMAT plans were recalculated for each case. Plans were also reoptimised on these data sets, and recalculated. Optimisation error was assessed through equivalent uniform dose (EUD) differences. Additionally, point dose comparison, dose volume histogram (DVH) analysis and gamma analysis of dose were used to assess dose calculation error. The dose calculation error on average was an increase in EUD whereas the optimisation error on average was a reduction in EUD compared to the original plan for all datasets aside from the bone only override dataset where bone was set to 1.61g/cm3. For the optimisation error, the largest mean absolute error (MAE) was 1.88Gy EUD for the PTV, and 2.21Gy EUD for the brainstem, for the reoptimisation completed on the air only overridden dataset, and recalculated on the original. Bulk density corrections for bone and air provide dose calculations within 3% of the original treatment plans. Optimisation errors have the potential to be greater than dose calculation errors if incorrect density corrections are utilized. Electron density correction using a bulk density approach achieves dose calculation uncertainties within 3%, however more advanced approaches, such as a voxel based approach, may improve accuracy and should be considered.

Integration of radiobiological modeling and indices in comparative plan evaluation: A study comparing VMAT and 3D-CRT in patients with NSCLC

PurposeThe purpose of this article was to generate an algorithm to calculate radiobiological endpoints and composite indices and use them to compare volumetric modulated arc therapy (VMAT) and 3-dimensional conformal radiation therapy (3D-CRT) techniques in patients with locally advanced non-small cell lung cancer. Methods and materialsThe study included 25 patients with locally advanced non-small cell lung cancer treated with 3D-CRT at our center between January 1, 2010, and December 31, 2014. The planner generated VMAT plans using clones of the original computed tomography scans and regions of interest volumes, which did not include the original 3D plans. Both 3D-CRT and VMAT plans were generated using the same dose-volume constraint worksheet. The dose-volume histogram parameters for planning target volume and relevant organs at risk (OAR) were reviewed. The calculation engine was written in the R programming language; the user interface was developed with the “shiny" R Web library. Dose-volume histogram data were imported into the calculation engine and tumor control probability (TCP), normal tissue complication probability (NTCP), composite cardiopulmonary toxicity index (CPTI), morbidity index: MI = ∑j = 1#ofrelevantOARs(wj ∗ NTCPj), uncomplicated TCP (UTCP=TCP∗∏k=1#ofOARs1−NTCPK100, and therapeutic gain (TG): ie, TG = TCP ∗ (100 − MI) were calculated. ResultsTCP was better with 3D-CRT (12.62% vs 11.71%, P < .001), whereas VMAT demonstrated superior NTCP esophagus (4.45% vs 7.39%, P = .02). NTCP spinal cord (0.001% vs 0.009%, P = .001), and NTCP heart/perfusion defect (44.57% vs 56.42%, P = .016). There was no difference in NTCP lung (6.27% vs 7.62%, P = .221) and NTCP heart/pericarditis (0.001% vs 0.15%, P = .129) between 2 techniques. VMAT showed substantial improvement in morbidity index (11.06% vs. 14.31%, P = 0.01), CPTI (47.59% vs 59.41%, P = .03), TG (P = .035), and trend toward superiority in UTCP (5.89 vs 4.75, P=.057). ConclusionThe study highlights the utility of the radiobiological algorithm and summary indices in comparative plan evaluation and demonstrates benefits of VMAT over 3D-CRT.

Construction of mouse phantoms from segmented CT scan data for radiation dosimetry studies

We present the complete construction methodology for an anatomically accurate mouse phantom made using materials which mimic the characteristics of tissue, lung, and bone for radiation dosimetry studies. Phantoms were constructed using 2 mm thick slices of tissue equivalent material which was precision machined to clear regions for insertion of lung and bone equivalent material where appropriate. Images obtained using a 3D computed tomography (CT) scan clearly indicate regions of tissue, lung, and bone that match their position within the original mouse CT scan. Additionally, radiographic films are used with the phantom to demonstrate dose mapping capabilities. The construction methodology presented here can be quickly and easily adapted to create a phantom of any specific small animal given a segmented CT scan of the animal. These physical phantoms are a useful tool to examine individual organ dose and dosimetry within mouse systems that are complicated by density inhomogeneity due to bone and lung regions.

Feasibility of Merging Sacroiliac Fusion Computed Tomography Preoperative Planning to Original Computed Tomography Scans

Background: Spinal merging has been done with preoperative CT scans merged with fluoroscopic imaging. However, preoperative and intraoperative CT scans have not been merged for the purposes of transferring a plan. If spine surgeons can create the optimal surgical pathway before surgery and then merge the pre-planned pathway onto the real patient CT scan in the operating room, the result will be efficiency and better time management for the spine surgeon. Our objective was to analyze the feasibility of merging computed tomography (CT) preplanned with sacroiliac fusion iliosacral screw trajectories with the original CT scans. Methods: Twenty-seven patients were identified from the institutional Faculty Group Practice who had undergone CT scans in the pelvic region. The pelvic CT scans were transferred to Medtronic's StealthStation ® , which allowed for analysis in coronal, sagittal, and axial sections. Three trajectories were mapped out, representing three working pathways for fixation. The position of the trajectories was localized by measuring their distance from anatomic landmarks. The CT scan with the planned trajectories was merged with the original CT scan. The same measurements made pre-merge were performed post-merge to check for the accuracy of the merging process. Results: Post-merge measurements were nearly identical to pre-merge measurements, with a maximum difference of ± 0.1 mm. No significant differences existed between pre-merge and post-merge measurements for any of the trajectories Conclusion: The merging of preplanned trajectory CT scans with intraoperative CT scans is a highly accurate technique. Clinical relevance: If this technique can be replicated by merging actual pre-operative CT scans with actual intra- operative scans, it could save time and expense. This would benefit the patient because time under anesthesia would be reduced.

Computed Tomography Findings in Patients With Pediatric Blunt Renal Trauma in Whom Expectant (Nonoperative) Management Failed

To determine whether the features on a computed tomography (CT) scan can predict the need for urologic intervention in a pediatric blunt renal trauma population initially treated with expectant management. A review of a prospective database of pediatric patients sustaining renal trauma from 1991 to 2003 was performed. The data reviewed included the mechanism of injury, injury grade, CT findings, operative intervention, and complications. Parametric statistical analysis was used to compare the CT findings and outcomes. A total of 72 children presented with blunt renal injury, of whom 61 met the study criteria. Of the 61 patients, 50 had grade I-III, 10 had grade IV, and 1 had grade V injuries. No children with grade I-III injuries required operative intervention. Of the 10 patients with grade IV injuries, 4 had medial contrast extravasation from the collecting system on their original CT scan, 3 of whom required intervention. Intervention initially consisted of delayed endoscopic procedures at 3, 9, and 33 days after injury. All 3 patients (100%) developed complications in their management, and 2 (66%) required open surgical intervention. The 1 patient with grade V injury underwent nephrectomy because of hemodynamic instability. Grade IV renal injuries with medial contrast extravasation are associated with urologic intervention at greater rates than those without extravasation. Delayed treatment of this finding could be associated with greater than expected complication rates and renal loss, and early/aggressive treatment should be considered. This knowledge could improve the specificity of "expectant" nonoperative management of pediatric renal injury.

Whole Brain Radiotherapy With Hippocampal Avoidance and Simultaneously Integrated Brain Metastases Boost: A Planning Study

To evaluate the feasibility of using tomotherapy to deliver whole brain radiotherapy with hippocampal avoidance, hypothesized to reduce the risk of memory function decline, and simultaneously integrated boost to brain metastases to improve intracranial tumor control. Ten patients treated with radiosurgery and whole brain radiotherapy underwent repeat planning using tomotherapy with the original computed tomography scans and magnetic resonance imaging-computed tomography fusion-defined target and normal structure contours. The individually contoured hippocampus was used as a dose-limiting structure (<6 Gy); the whole brain dose was prescribed at 32.25 Gy to 95% in 15 fractions, and the simultaneous boost doses to individual brain metastases were 63 Gy to lesions >or=2.0 cm in the maximal diameter and 70.8 Gy to lesions <2.0 cm. The plans were generated with a field width (FW) of 2.5 cm and, in 5 patients, with a FW of 1.0 cm. The plans were compared regarding conformation number, prescription isodose/target volume ratio, target coverage, homogeneity index, and mean normalized total dose. A 1.0-cm FW compared with a 2.5-cm FW significantly improved the dose distribution. The mean conformation number improved from 0.55 +/- 0.16 to 0.60 +/- 0.13. Whole brain homogeneity improved by 32% (p <0.001). The mean normalized total dose to the hippocampus was 5.9 +/- 1.3 Gy(2) and 5.8 +/- 1.9 Gy(2) for 2.5- and 1.0-cm FW, respectively. The mean treatment delivery time for the 2.5- and 1.0-cm FW plans was 10.2 +/- 1.0 and 21.8 +/- 1.8 min, respectively. Composite tomotherapy plans achieved three objectives: homogeneous whole brain dose distribution equivalent to conventional whole brain radiotherapy; conformal hippocampal avoidance; and radiosurgically equivalent dose distributions to individual metastases.

Retroperitoneal imaging with third and fourth generation computed axial tomography in clinical stage I nonseminomatous germ cell tumors

Objectives. To examine the accuracy rate of abdominal staging using thirdand fourth generation computed tomography (CT) scanning in clinical Stage I testicular nonseminoma patients. Methods. Between January 1985 and August 1993, 57 patients presented to ourcenter with clinical Stage I testicular nonseminoma. Retroperitoneal computed tomographic staging studies were interpreted to be normal preoperatively in the entire group. In addition, tumor marker values were normal or returned to normal postorchiectomy within the appropriate half-life intervals. All patients were subjected to radical or modified retroperitoneal lymph node dissection (19% and 72%, respectively). Original abdominal CT scans (preretroperitoneal lymph node dissection) were available for blinded retrospective re-review in 16 cases (7 pathologic Stage II, 9 pathologic Stage I). Results. Nineteen of 57 (33%) patients were upstaged at surgery including 6 patients (11%) who demonstrated II B volume disease. Third and fourth generation CT scanning of the retroperitoneum yielded a 66% accuracy rate in this population. Six out of 7 pathologic Stage II pre-lymph node dissection abdominal CT scans that were available for blinded re-review revealed nonpathologic nodes by size criteria in the primary landing zone for the corresponding original tumor. Conclusions. Our data suggests that for clinical Stage I nonseminoma in the 1985 to1993 era, undue reliance was placed on a less than ideal staging test. The 33% false-negative rate reported showed no improvement over earlier reports and reaffirms concern for relying solely on third or fourth generation CT imaging of the retroperitoneum in the staging of clinical Stage I nonseminomatous germ cell tumor (NSGCT) patients. The presence of any number of lymph nodes in the expected primary landing zone, regardless of size, should raise serious suspicion for occult nodal disease.

Acute bilateral ureteric obstruction complicating pelvic fracture

A 23-year-old male motorcyclist sustained a Tile type B (open book) pelvic fracture (Figure I) in a direct collision with a car. He was admitted to hospital and resuscitated according to ATLS guidelines. In addition to 24 units of blood, an external fixator was applied to his pelvis to gain haemodynamic stability; however, incomplete reduction was gained (open anatomical reduction and fixation was performed later). Computed tomography (CT) excluded other abdominal injuries but revealed a haematoma filling his pelvis (Figure 2). An abdominal ultrasound scan performed 20 h later revealed bilateral hydronephrosis (not present on the original CT scan), confirmed by contrast study performed after insertion of the nephrostomy tube, and urgent cystogram (Figwe3) showed severe displacement and compression of an intact bladder. The radiograph shows the contrast in the bladder resembling a hot-air balloon rising out of the pelvis. On admission his urea level was 6.5 mmol/l and creatinine was 80 mmol/l. After 24 h his urea was 14.1 and creatinine 229, with a potassium level of 6.7 mmol/l and a metabolic acidosis. He

Prediction of outcome in senile dementia--a computed tomography study.

Forty patients suffering from senile dementia who had been subjected to detailed clinical and psychological assessment and computed tomography (CT) were followed up for a mean period of 28.78 months. All but one were traced. The deceased (27) were compared with the survivors (12). All males had died at follow-up. The mean age was not statistically significant, but the survivors differed significantly from the deceased in having performed better on a number of clinical and psychological tests, particularly those involving speech functions and constructional ability. Measures of ventricular size and cortical atrophy were not of predictive value, but a new technique of measuring radiological density showed that this was significantly lower in the right parietal region in the original CT scans of those who subsequently died. The study confirms that clinical involvement of the parietal lobes is an indication of poor prognosis and reports the first radiological support for this view. It also suggests that a more directly quantitative approach to computed tomography may yield results which are more useful than those obtained from visual reconstructions.