All-star approach to a small medical imaging dataset: combined deep, transfer, and classical machine learning approaches for the determination of radial head fractures.

BACKGROUND: Radial head fractures are often evaluated in emergency departments and can easily be missed. Automated or semi-automated detection methods that help physicians may be valuable regarding the high miss rate. PURPOSE: To evaluate the accuracy of combined deep, transfer, and classical machine learning approaches on a small dataset for determination of radial head fractures. MATERIAL AND METHODS: A total of 48 patients with radial head fracture and 56 patients without fracture on elbow radiographs were retrospectively evaluated. The input images were obtained by cropping anteroposterior elbow radiographs around a center-point on the radial head. For fracture determination, an algorithm based on feature extraction using distinct prototypes of pretrained networks (VGG16, ResNet50, InceptionV3, MobileNetV2) representing four different approaches was developed. Reduction of feature space dimensions, feeding the most relevant features, and development of ensemble of classifiers were utilized. RESULTS: The algorithm with the best performance consisted of preprocessing the input, computation of global maximum and global mean outputs of four distinct pretrained networks, dimensionality reduction by applying univariate and ensemble feature selectors, and applying Support Vector Machines and Random Forest classifiers to the transformed and reduced dataset. A maximum accuracy of 90% with MobileNetV2 pretrained features was reached for fracture determination with a small sample size. CONCLUSION: Radial head fractures can be determined with a combined approach and limitations of the small sample size can be overcome by utilizing pretrained deep networks with classical machine learning methods.

Variations ignored in routine wrist MRI reports: prevalence of the median nerve anatomical variations and persistent median artery.

BACKGROUND: It is vital to know the anatomical variations of the wrist to avoid iatrogenic injuries during carpal tunnel (CT) surgery. PURPOSE: To determine the anatomical variations of the median nerve (MN) and the prevalence of persistent median artery (PMA) on wrist magnetic resonance imaging (MRI). MATERIAL AND METHODS: A total of 300 wrists evaluated by MRI during 2013-2015 were retrospectively identified. While branching of the MN distal to the CT is accepted as the normal anatomy, proximal to the tunnel and within the tunnel were considered as variations. The prevalence of PMA was also evaluated. The patients were assigned to groups according to age, gender, and wrist side and compared to determine whether there was any significant difference in terms of these variations. All evaluations were assessed with the shared decision of a musculoskeletal radiologist and a radiology resident. RESULTS: Of the 300 wrists, 38 (12.7%) and 34 (11.3%) had a bifid MN proximal to the CT and within the CT, respectively. Only one nerve trifurcation was seen within the CT. The MN exhibited branching distal to the CT in 227 (76%) patients. PMA was observed in 44 (14.7%) patients. Of the 44 PMA cases, 28 (63.6%) also had a coexisting MN variation. There was no significant difference in the prevalence of MN variations and PMA in the subgroups (P > 0.05). CONCLUSION: Nearly one in four patients (24.4%) have MN variations and 14.8% had PMA. Preoperative evaluation of these common anatomical variations with MRI will be protective against postoperative complications of CT surgery.

Investigation of ligament, bone, synovial and plica pathologies accompanied by common extensor tendon in patients with lateral overuse syndrome of the elbow using magnetic resonance imaging.

BACKGROUND: Tendinosis in the common extensor tendon and accompanying ligament, bone, and plica abnormalities can be observed on magnetic resonance imaging (MRI). PURPOSE: To determine whether there is a difference between accompanying abnormalities according to the degree of common extensor tendon injury. MATERIAL AND METHODS: Patients who underwent 1.5-T MRI tests with a prediagnosis of lateral overuse syndrome were retrospectively reviewed, and 56 patients who had an injury in the common extensor tendon (CET) were included. The degree of tendon and ligament injury, muscle signal change, bone marrow signal change, presence of joint effusion, and morphological features in the presence of plica were evaluated via MRI examinations of the elbow. RESULTS: Overall, 32, 16, and eight patients had mild, moderate, and severe CET damage, respectively. As the severity of CET damage increased, the presence of joint effusion, and the presence and degree of damage to the lateral ulnar collateral ligament (LUCL) and radial collateral ligament (RCL) increased. The radiohumeral (RH) plica area was significantly larger in the group with mild CET damage. There was no statistically significant correlation between the severity of CET damage and the end of RH plica with a blind-end, coverage of one-third or more of the radius, its signal, thickness, and presence of olecranon fold. CONCLUSION: As the severity of CET injury increases, damage to the LUCL, RCL, and the presence of effusion in the joint increases. RH plica should be evaluated in terms of concomitant pathology in patients with mild CET injuries on MRI.

The effect of ulnar variance on scapholunate and capitolunate angles.

Background Ulnar variance (UV), which is measured using anteroposterior (AP) X-rays, is associated with the development of multiple wrist pathologies. The scapholunate angle (SLA) and capitolunate angle (CLA) are measured using lateral X-rays, and these can be used in the diagnosis of intercalated segmental instability. Purpose To determine the effect of UV on SLA and CLA. Material and Methods A total of 140 patients (73 women, 67 men; mean age, 37.8 ± 14.6 years; 95% confidence interval [CI], 35.4-40.3) were included in the study. Participants were excluded if they presented with fractures or malunions, any arthritic conditions, avascular necrosis, congenital deformities, or bone and soft tissue tumors. UV, SLA, and CLA were measured using AP and lateral wrist X-rays. Results Patients were grouped as positive, neutral, and negative UV. There was no statistically significant difference in mean ages, sex, and sides (left or right) between the UV groups ( P > 0.05). In addition, there was no statistically significant difference in the mean values of SLA and CLA between the UV groups ( P > 0.05). UV showed no statistically significant association with SLA and CLA (r = -0.064; P = 0.455, and r = 0.059; P = 0.485, respectively). However, there was a statistically significant association between SLA and CLA (r = -0.482; P < 0.001). Conclusion There is higher prevalence of neutral UV and no correlation between UV with respect to age and sex of patients in the local Turkish population. SLA and CLA are not affected by UV. There is a negative correlation between SLA and CLA.

The anatomic variations of the posterior talofibular ligament associated with os trigonum and pathologies of related structures.

PURPOSE: The aim of this study is to investigate the detailed anatomy of the posterior talofibular ligament (PTFL) on MR images in patients with os trigonum. We also evaluated the pathological conditions of the PTFL, anterior talofibular ligament (ATFL), flexor hallucis longus (FHL) tendon, talus and os trigonum. METHODS: Ankle MRIs of 70 patients with os trigonum (study group) and 70 patients without it (control group) were reviewed for the anatomy of the anterior and posterior fibers of PTFL. The prevalence of PTFL and ATFL pathologies was also compared between two groups. Additionally FHL tenosynovitis and osseous pathologies were evaluated. RESULTS: The posterior fibers inserted into the lateral tubercule of the posterior process of the talus in the control group whereas if an os trigonum was present, the posterior fibers of PTFL were inserted only into the os trigonum. The origins of anterior and posterior fibers were the medial surface of the lateral malleolus and the insertion of the anterior fibers was lateral surface of the talus posterior to the lateral malleolar facet in both groups. There was a significant association between an abnormal PTFL, ATFL and the presence of os trigonum. FHL tenosynovitis was higher in the study group but it did not meet the statistical significance. The most common pathology of the talus and os trigonum was subchondral edema along the synchondrosis. CONCLUSIONS: In patients with os trigonum, the posterior fibers of the PTFL were inserted herein. In the case of an os trigonum signal alterations of ligaments were more common, which may reflect chronic instability.

Searching for the cheese: does ethmoid pneumatization affect the location of the anterior ethmoid artery?

Endoscopic sinus surgery is a widely used technique in otolaryngologic practice. To avoid complications, the locations of important anatomical structures, such as the anterior ethmoid artery (AEA), should be determined preoperatively. We want to evaluate the effect of ethmoid cavity pneumatization on the location of the AEA and to determine consistent landmark(s) for locating the AEA. 524 consecutive patients undergoing sinus CT scans between February and October 2012 were reviewed retrospectively. After the exclusion criteria were applied, 150 CT scans (300 sides) were selected for the study. A statistically significant positive correlation was found between ethmoid pneumatization and the distance of the AEA to the attachment of the inferior turbinate to the lateral nasal wall (Spearman's rho = 0.305; p < 0.001). Likewise, a statistically significant positive correlation was found between ethmoid pneumatization and the distance between the AEA and the frontonasal junction (Spearman's rho = 0.219; p < 0.001). We found that the artery was located mostly between the second and third lamellae [n 211 (71 %) cases]. There was no statistically significant correlation between ethmoid pneumatization and AEA location in terms of the lamellae. Increased ethmoid volume increases the distance of the AEA from the frontonasal junction and the lateral attachment of the inferior turbinate. However, increased pneumatization of the ethmoid cavity did not affect AEA localization in terms of the lamellae. Based on our findings, we suggest that using the lamellae to locate the AEA is reliable.

Voxel level dense prediction of acute stroke territory in DWI using deep learning segmentation models and image enhancement strategies.

PURPOSE: To build a stroke territory classifier model in DWI by designing the problem as a multiclass segmentation task by defining each stroke territory as distinct segmentation targets and leveraging the guidance of voxel wise dense predictions. MATERIALS AND METHODS: Retrospective analysis of DWI images of 218 consecutive acute anterior or posterior ischemic stroke patients examined between January 2017 to April 2020 in a single center was carried out. Each stroke area was defined as distinct segmentation target with different class labels. U-Net based network was trained followed by majority voting of the voxel wise predictions of the model to transform them into patient level stroke territory classes. Effects of bias field correction and registration to a common space were explored. RESULTS: Of the 218 patients included in this study, 141 (65%) were anterior stroke, and 77 were posterior stroke (35%) whereas 117 (53%) were male and 101 (47%) were female. The model built with original images reached 0.77 accuracy, while the model built with N4 bias corrected images reached 0.80 and the model built with images which were N4 bias corrected and then registered into a common space reached 0.83 accuracy values. CONCLUSION: Voxel wise dense prediction coupled with bias field correction to eliminate artificial signal increase and registration to a common space help models for better performance than using original images. Knowing the properties of target domain while designing deep learning models is important for the overall success of these models.

Deep Learning Classification of Ischemic Stroke Territory on Diffusion-Weighted MRI: Added Value of Augmenting the Input with Image Transformations.

Our primary aim with this study was to build a patient-level classifier for stroke territory in DWI using AI to facilitate fast triage of stroke to a dedicated stroke center. A retrospective collection of DWI images of 271 and 122 consecutive acute ischemic stroke patients from two centers was carried out. Pretrained MobileNetV2 and EfficientNetB0 architectures were used to classify territorial subtypes as middle cerebral artery, posterior circulation, or watershed infarcts along with normal slices. Various input combinations using edge maps, thresholding, and hard attention versions were explored. The effect of augmenting the three-channel inputs of pre-trained models on classification performance was analyzed. ROC analyses and confusion matrix-derived performance metrics of the models were reported. Of the 271 patients included in this study, 151 (55.7%) were male and 120 (44.3%) were female. One hundred twenty-nine patients had MCA (47.6%), 65 patients had posterior circulation (24%), and 77 patients had watershed (28.0%) infarcts for center 1. Of the 122 patients from center 2, 78 (64%) were male and 44 (34%) were female. Fifty-two patients (43%) had MCA, 51 patients had posterior circulation (42%), and 19 (15%) patients had watershed infarcts. The Mobile-Crop model had the best performance with 0.95 accuracy and a 0.91 mean f1 score for slice-wise classification and 0.88 accuracy on external test sets, along with a 0.92 mean AUC. In conclusion, modified pre-trained models may be augmented with the transformation of images to provide a more accurate classification of affected territory by stroke in DWI.

Detection and vascular territorial classification of stroke on diffusion-weighted MRI by deep learning.

PURPOSE: Rapid detection and vascular territorial classification of stroke enable the determination of the most appropriate treatment. In this study, we aimed to investigate the performance of convolutional neural network (CNN) models in the detection and vascular territorial classification of stroke on diffusion-weighted images (DWI). METHODS: DWI of 421 cases (271 acute ischemic stroke patients and 150 cases without any ischemia findings on DWI) obtained between January 2017 to April 2020 were reviewed. We created two custom datasets. A stroke detection dataset was created with 1800 slices (900 S and 900 normal) consisting of 1400 for training, 200 for validation, 200 for test. A vascular territorial type dataset was created with 1717 slices (883 middle cerebral artery stroke, 416 posterior circulatory stroke, and 418 watershed stroke) consisting of 1117 slices for training, 300 for validation, 300 for test. A transfer learning approach based on MobileNetV2 and EfficientNet-B0 CNN architecture was used. The performance of the models was evaluated. RESULTS: Modified MobileNetV2 and EfficientNet-B0 models achieved 96% (κ: 0.92) and 93% (κ: 0.86) accuracy in stroke detection, respectively. In vascular territorial classification of stroke as middle cerebral artery, posterior circulation, or watershed infarction, an accuracy of 93% (κ: 0.895) was achieved with modified MobileNetV2 model and 87% (κ: 0.805) with modified EfficientNet-B0 CNN model. CONCLUSION: Transfer learning approach with custom top CNN models achieve sufficiently high performance for both the detection of ischemic stroke and the classification of its vascular territorial type on DWI.

Torsion of Wandering Spleen: Importance of Splenic Density and Liver-to- Spleen Attenuation Ratio on CT.

BACKGROUND: Wandering spleen (WS) is a rare clinical condition which may cause fatal complication like torsion with subsequent infarction. Determination of splenic parenchyma viability is very important in deciding whether splenopexy rather than splenectomy is an option. Contrast- enhanced computed tomography (CECT) is important for the diagnosis of WS and assessment of the viability of spleen. DISCUSSION: We reviewed the CT studies of four cases with WS. We measured the mean splenic and liver density and calculated liver-to-spleen attenuation ratio (LSAR). We also assessed the CT findings for each patient. Mean splenic density was measured as 40.77 Hounsfield Unit (HU) in cases with infarction, 127.1 HU in case without infarction. LSAR was calculated as 2.55 in cases with infarction, 0.99 in case without infarction. We detected whirlpool sign, intraperitoneal free fluid, splenic arterial enhancement in all patient, parenchymal and splenic vein enhancement in one patient without infarction, fat rim sign in three patients with infarction, capsular rim sign in one patient with infarction. CONCLUSION: CECT should be obtained for the diagnosis of WS and assessment of the viability of spleen. CECT could suggest the diagnosis of infarction of the spleen with following findings; absence of parenchymal enhancement, very low density of spleen (<45 HU), and LSAR which is greater than 2.

The association of patellar maltracking with infrapatellar fat pad edema and chondromalacia patella: A quantitative morphological magnetic resonance imaging analysis.

OBJECTIVES: This study aims to investigate the relationship of patellofemoral joint morphology with infrapatellar fat pad edema and chondromalacia patella (CP) using patellar maltracking parameters on magnetic resonance imaging (MRI). PATIENTS AND METHODS: Between January 2010 and January 2013, 50 patients with edema in the superolateral portion of the infrapatellar fat pad (the study group) and control group (n=50) with a normal infrapatellar fat pad were identified on MRI to compare with regard to five patellar maltracking parameters retrospectively. These parameters were trochlear depth, the trochlear sulcus angle (TSA), patellar translation, the lateral patellofemoral angle (PFA), and the Insall-Salvati ratio. The relationship between patellar maltracking and the CP was also evaluated using the same parameters. RESULTS: In the study group, the Insall-Salvati index and TSA were significantly higher (p=0.001), and the trochlear depth and PFA were low (p=0.001, p=0.01), while patellar translation showed no difference (p=0.957). In the CP group, the Insall-Salvati index and TSA were significantly high (p=0.001), the trochlear depth was low (p=0.001). No statistically significant difference was found in PFA and patellar translation (p=0.292, p=0.446). CONCLUSION: Our study results suggest that edema in the superolateral portion of infrapatellar fat pad and CP are associated with patellar maltracking.

Venous flow volume measured by duplex ultrasound can be used as an indicator of impaired tissue perfusion in patients with peripheral arterial disease.

AIM: In this study, we aimed to investigate the arterial and venous flow volume rate (FV) in order to determine the tissue perfusion using duplex ultrasonography (DU). We hypothesized that FV provides reliable information regarding tissue perfusion in patients with peripheral arterial disease (PAD). MATERIAL AND METHODS: The study comprised 38 patients (72 legs) with PAD. In all patients, common femoral, popliteal, anterior tibial, posterior tibial arteries and veins were examined with DU. Measurements were obtained in the supine position with 15 cm elevation of the foot to neutralise central venous pressure. The diameter, blood flow velocity, and FV of arteries and veins were measured for each patient. RESULTS: The FV of the common femoral artery and vein (p = 0.001), popliteal artery and vein (p=0.003), and posterior tibial artery and vein (p = 0.008) had statistically significant differences. However, there was no statistically significant difference between the FV of the anterior tibial vein and artery (p = 0.408). The mean FV values of all veins were significantly lower than those of homonymous arteries in patients with PAD. CONCLUSIONS: Our study showed that venous FV measured by DU can be used as an indicator of impaired tissue perfusion in patients with PAD.

New technique for the treatment of bilateral vocal cord paralysis: Vocal and ventricular fold lateralization using crossing sutures with thyroplasty technique.

CONCLUSION: All patients treated with this new lateralization technique had a good quality of life and no dyspnea at rest or upon exertion. We believe that this technique is an important addition to the many surgical techniques for the treatment of bilateral vocal cord paralysis (BVCP). OBJECTIVES: Most techniques used in the treatment of BVCP result in a prolapse of the laryngeal soft tissues into the endolarynx owing to Bernoulli's principle. We have developed a new lateralization technique to more effectively prevent this prolapse. METHODS: The lateralization was initially tested on six cadaver larynges before being performed in five clinical cases with BVCP, who suffered from dyspnea at rest. Average follow-up was 17.6 months. As in type 1 thyroplasty, a rectangular piece of cartilage was excised horizontally from the thyroid lamina and placed on the defect in the vertical plane. Then, the vocal and ventricular folds were lateralized with crossing sutures, particularly in the posterior region. RESULTS: Dyspnea was eliminated postoperatively. All patients were successfully decannulated. Postoperative voice quality was socially acceptable. Airways were improved postoperatively, as evaluated via fiberoptic laryngoscopy and CT. The mean preoperative and postoperative rima openings were 1.3 mm (range 0.5-2.6) and 6.4 mm (range 3.4-8.1), respectively.