Are deep learning classification results obtained on CT scans fair and interpretable?

Following the great success of various deep learning methods in image and object classification, the biomedical image processing society is also overwhelmed with their applications to various automatic diagnosis cases. Unfortunately, most of the deep learning-based classification attempts in the literature solely focus on the aim of extreme accuracy scores, without considering interpretability, or patient-wise separation of training and test data. For example, most lung nodule classification papers using deep learning randomly shuffle data and split it into training, validation, and test sets, causing certain images from the Computed Tomography (CT) scan of a person to be in the training set, while other images of the same person to be in the validation or testing image sets. This can result in reporting misleading accuracy rates and the learning of irrelevant features, ultimately reducing the real-life usability of these models. When the deep neural networks trained on the traditional, unfair data shuffling method are challenged with new patient images, it is observed that the trained models perform poorly. In contrast, deep neural networks trained with strict patient-level separation maintain their accuracy rates even when new patient images are tested. Heat map visualizations of the activations of the deep neural networks trained with strict patient-level separation indicate a higher degree of focus on the relevant nodules. We argue that the research question posed in the title has a positive answer only if the deep neural networks are trained with images of patients that are strictly isolated from the validation and testing patient sets.

Correlation of 3-T MRI and CT findings with patient symptoms and treatment outcome in radiofrequency ablation of osteoid osteoma.

OBJECTIVE: The aim of this prospective study was to evaluate pre- and post-treatment MRI and CT findings of osteoid osteoma (OO) patients treated with radiofrequency thermo-ablation (RFTA) and to compare these findings with visual analog scale (VAS) scores. METHODS: Sixteen patients (4 females and 12 males; mean age of 18.87 ± 8.75 years (range: 8-37)) with OO were examined with CT and MRI, at baseline and at an average of 3 months following the procedure. On pre- and post-procedural CT and MRIs, OO-related findings were recorded. Treatment success was evaluated with VAS scores. RESULTS: Baseline VAS scores were 8 or 9 and follow-up scores were 0 or 1, indicating no early recurrences. Nidus diameters decreased significantly after the procedure (p = 0.027, p = 0.002, and p = 0.002; and p = 0.001, p = 0.001, p = 0.001 for AP, ML and CC nidus diameters for CT and MRI, respectively). The mean nidus volume were significantly decreased after the procedure (p = 0.001, for CT and MRI). On post-procedural images, cortical thickening, the signal intensity and contrast enhancement of the nidus and the extent of periostitis were significantly decreased (p = 0.019, p = 0.001, p = 0.001 and p = 0.034, respectively). There was no significant change in nidus calcification, perinidal cortical and intramedullary sclerosis, periosteal reaction, bone deformity, bone marrow and soft tissue edema, joint effusion and synovitis after the procedure (p = 0.253, p = 0.062, p = 0.245, p = 1, p = 1, p = 0.429, p = 0.371, p = 0.625, p = 1). CONCLUSION: Although the changes in imaging findings may be helpful in early follow-up of OO patients treated with RFTA, these changes alone cannot be used with accuracy in predicting treatment response. LEVEL OF EVIDENCE: Level IV, Therapeutic Study.

Prevalence of elastofibroma dorsi on CT: Is it really an uncommon entity?

OBJECTIVE: The aim of this study was to determine the prevalence, demographic data of elastofibroma dorsi (ED) in adult population who had undergone chest CT examination and to discuss clinical, and radiological presentations, and treatment options of ED. METHODS: We retrospectively reviewed 4074 chest CT examinations for ED from July 2014 to April 2015. Lesion size, side, and patient demographics were analyzed for positive cases of ED. The initial radiology reports of patients with ED were also reviewed. RESULTS: Of the 4074 patients, 111 patients (2.73%) (77 women and 34 men; mean age: 68.2 years; range: 35-91 years) had a total of 168 ED. The females had a 1.96 -fold higher prevalence of ED than the males (OR, 1.96; 95% CI, 1.481-2.59). The mean lesion thickness was found to be significantly greater in the female patients compared with the male patients (p = 0.001). The prevalence of the disease was estimated to be 4.98 times higher in patients aged 65 years or older (CI 95%, 3.25-7.36). In 111 ED patients, the lesions were only noted in 9 patients' initial radiology report. CONCLUSION: Here, we present a prevalence study with the largest population in the literature concerning ED. Our study shows that ED is not as uncommon as previously thought and should be especially suspected in females and older age groups. LEVEL OF EVIDENCE: Level IV, Diagnostic Study.

Can neck pain be an initial symptom of rheumatoid arthritis? A case report and literature review.

Although rheumatoid arthritis (RA) is most commonly associated with peripheral joints, cervical spine involvement can be seen in almost 80% of patients in the presence of long-term disease, joint erosion, and risk factors such as male sex and rheumatoid factor positivity. It is very rare to have cervical involvement in the initial period of RA. If a patient has isolated cervical spine involvement without peripheral arthritis, it is highly likely that inappropriate investigations and delayed treatment may occur. Any damage that occurs in cervical spine may cause symptoms varying from slight instability to atlantoaxial subluxation, spinal cord and brain stem compression and even death. Therefore, physician should be aware that there may be isolated cervical involvement, albeit rare, in patients with RA. In this report, we presented a case of RA presenting with cervical spine involvement without peripheral arthritis to underline the importance of this kind of involvement in clinical practice. We also briefly reviewed other cases similar to ours in light of literature.

Definition of a quantitative measurement method for acetabular version in a plain radiograph in the healthy adult hip.

OBJECTIVES: This study aims to define a quantitative measurement method for acetabular version in a standard anteroposterior hip radiograph, assess the intraobserver and interobserver agreements of this method, and compare it with the gold standard computed tomography (CT). PATIENTS AND METHODS: Anteroposterior standard hip radiographs and simultaneously taken transverse acetabular CT sections of 78 hips of 39 patients (10 males, 29 females; mean age 60 years; range 40 to 81 years) were used in the study. In standard anteroposterior hip radiographs, "acetabular anterior wall line" was identified as the line between the most lateral edge of the acetabulum and the inferolateral edge of the teardrop. "Acetabular posterior wall line" was identified as the line between the most lateral edge of the subchondral sclerosis and the outmost point of acetabulum posterior lunate surface sclerosis. To assess the reliability of this technique, mentioned angles in 78 hips were measured by two authors independently two weeks apart. Direct radiographic values were compared with the acetabular version measurement values in CT examination. RESULTS: Mean acetabular version angles of 78 hips in plain radiographs and CT were 18.0° (9-25°) and 17.2° (12-25°), respectively. Mean intraobserver measurement differences were 1.3° (0-5°) and 1.5° (0-6°). Mean interobserver measurement difference was 1.4° (0-5°). The mean difference between plain radiography measurements and CT measurements was 2.5° (0-6°). A significant correlation was detected between plain radiographic measurements and CT measurements. CONCLUSION: By this quantitative method, acetabular morphology may be measured less invasively, easily, quickly and reliably in plain radiograph in transverse plane.

A novel KIF7 mutation in two affected siblings with acrocallosal syndrome.

Acrocallosal syndrome (ACLS) is a rare genetic disorder typically characterized by craniofacial dysmorphism, agenesis, or hypoplasia of the corpus callosum, and duplication of the phalanges of halluces and/or the thumbs. ACLS is a recessive ciliopathy caused by mutations in KIF7. We identified a Turkish family who had a novel homozygous sequence change, c.2593-2A>C, located at the acceptor splice site of intron 12 of KIF7 (IVS12-2A>C). The present report will contribute towards further understanding of the genotype-phenotype correlation in ACLS caused by KIF7 mutations.

Variations in the popliteal artery branching in 342 patients studied with peripheral CT angiography using 64-MDCT.

PURPOSE: The objective of this study was to assess the utility of peripheral computed tomography angiography (CTA) using 64-multidetector computed tomography (64-MDCT) in the detection of variations in branching patterns of the popliteal artery. MATERIALS AND METHODS: The peripheral CTAs of 742 limbs in 342 patients who underwent peripheral CTA using 64-MDCT of the lower extremity due to various reasons were retrospectively reviewed. The anatomical variations in popliteal artery branching were assessed in 636 limbs from the same cohort of patients. RESULTS: Five hundred fifty-three (87 %) limbs had the usual branching pattern (type IA). Variations in the branching pattern of the popliteal artery were seen in 83 (13 %) limbs. The most common anatomical variation was trifurcation of the popliteal artery in 27 (4.2 %) limbs, with anterior tibial, posterior tibial, and peroneal arteries arising together with no true tibioperoneal trunk (type IB). CONCLUSION: The presence of variations in the anatomy of the popliteal artery could be of radiological and surgical importance in orthopedic interventions and various vascular surgery procedures. Peripheral CTA using MDCT is a suitable imaging modality for noninvasive evaluation of lower extremity arteries, and it may become the preferred method for evaluating variations in popliteal artery branching.

Comparison of the symptomatic and chondroprotective effects of glucosamine sulphate and exercise treatments in patients with knee osteoarthritis.

OBJECTIVES: The aim of the present study was to evaluate the symptomatic effects of glucosamine sulphate (GS) in comparison to the exercise therapy, as well effects on Magnetic Resonance Imaging (MRI) findings of cartilage loss in patients with knee osteoarthritis (OA). MATERIALS AND METHODS: Seventy patients with a diagnosis of knee OA were randomly divided into two groups. First group of patients (n=40) were treated with 1500 mg/day oral glucosamine sulphate and the second group (n=30) performed a home exercise program for a period of 6 months. RESULTS: Compared with baseline, significant improvements were observed in the visual analogue scale pain scores and the Western Ontario and McMaster Universities osteoarthritis index and 20-min walking time in both groups (p< 0.001); there were no significant differences between the two treatment groups. Upon the evaluation of joint cartilage thickness by MRI, an improvement was evident in only in the area of medial femoral condyle of patients who had performed a home exercise program (p< 0.05). CONCLUSION: The results of the present study suggest that both oral GS treatment and exercise therapy are efficient in the symptomatic treatment of patients with knee OA. However, exercise therapy found to have a better chondroprotective effect than oral GS treatment in this patient population.

Evaluation of Labral Pathology and Hip Articular Cartilage in Patients with Femoroacetabular Impingement (FAI): Comparison of Multidetector CT Arthrography and MR Arthrography.

BACKGROUND: To compare the multidetector computed tomography (MDCT) arthrography (CTa) and magnetic resonance (MR) arthrography (MRa) findings with surgical findings in patients with femoroacetabular impingement (FAI) and to evaluate the diagnostic performance of these methods. MATERIAL/METHODS: Labral pathology and articular cartilage were prospectively evaluated with MRa and CTa in 14 hips of 14 patients. The findings were evaluated by two musculoskeletal radiologists with 10 and 20 years of experience, respectively. Sensitivity, specificity, accuracy, and positive predictive value were determined using surgical findings as the standard of reference. RESULTS: While the disagreement between observers was recorded in two cases of labral tearing with MRa, there was a complete consensus with CTa. Disagreement between observers was found in four cases of femoral cartilage loss with both MRa and CTa. Disagreement was also recorded in only one case of acetabular cartilage loss with both methods. The percent sensitivity, specificity, and accuracy for correctly assessing the labral tearing were as follows for MRa/CTa, respectively: 100/100, 50/100, 86/100 (p<0.05). The same values for acetabular cartilage assessment were 89/56, 40/60, 71/71 (p>0.05) and for femoral cartilage assessment were 100/75, 90/70, 86/71 (p>0.05). Inter-observer reliability value showed excellent agreement for labral tearing with CTa (κ=1.0). Inter-observer agreement was substantial to excellent with regard to acetabular cartilage assessment with MRa and CTa (κ=0.76 for MRa and κ=0.86 for CTa). CONCLUSIONS: Inter-observer reliability with CTa is excellent for labral tearing assessment. CTa seems to have an equal sensitivity and a higher specificity than MRa for the detection of labral pathology. MRa is better, but not statistically significantly, in demonstrating acetabular and femoral cartilage pathology.

Imaging findings of an extradigital fibro-osseous pseudotumor.

Fibro-osseous pseudotumor (FOPT) is a rare and benign ossifying lesion. Described as the superficial variant of myositis ossificans (MO), this rare entity mostly occurs in the subcutaneous tissues of the digits. The FOPT clinicopathological features may mimic a variety of benign and malignant soft tissue lesions, and the diagnosis can be difficult when it arises in an unusual anatomic location. In this report we describe the clinical and radiological features of a case of an FOPT that involved the hypothenar region of the hand.

Inhalation therapy of calcitonin relieves osteoarthritis of the knee.

This study was conducted to determine if nasal salmon calcitonin has additional beneficial effects on clinical symptoms, serum NO, IL-1ß, matrix metalloproteinase 3, urinary C-terminal telopeptide type II collagen (CTX-II) levels and MRI findings in knee osteoarthritis (OA) when used concomitantly with exercise therapy. Fifty female patients with knee OA were randomized into two groups. The first group (n = 30) received 200 IU/day nasal salmon calcitonin and a home exercise program; the second group (n = 20) received a home exercise program for 6 months. Compared with baseline,while significant improvements were observed in visual analogue scale (VAS), WOMAC pain, physical function scores, 20-m walking time (P < 0.001) and WOMAC stiffness score (P = 0.041) in the first group, walking and resting VAS, and WOMAC physical function scores were improved (P = 0.029) in the second group after treatment. Significantly increased levels of serum NO and urinary CTX-II (P < 0.001) and significant improvements in the area of medial femoral condyle (P < 0.05) were noted only in the first group. There were significant differences in VAS activation values (P = 0.032) and NO levels (P < 0.001) in the favor of the first group. In conclusion, nasal salmon calcitonin may have possible chondroprotective effects besides its known effects on symptoms in patients with knee OA.

Recognition, assessment, and treatment of non-union after surgical fixation of fractures: emphasis on 3D CT.

For the post-operative patient, conventional axial computed tomography (CT) imaging and 2D multiplanar reconstruction are of limited value because of the beam-hardening artifact. However, three-dimensional (3D) CT imaging is an effective means of detecting subtle fracture healing, or confirming non-union, and for evaluating the integrity of metal hardware. In this article we emphasize the advantages of 3D CT imaging in the assessment and preoperative planning of non-union for patients who have been surgically treated for fractures.

Incomplete left-hand and complete right-hand pisiform and hamate coalition: multidetector computed tomography findings.

Pisiform-hamate coalition is a rare form of carpal coalition. Only 14 cases of pisiform-hamate coalition have been reported in the English-language literature. We present a case of asymptomatic bilateral pisiform-hamate coalition in a 16-year-old boy. We also review the embryology, pathogenesis, and clinical features of pisiform and hamate coalition, along with the associated multidetector computed tomography findings.

Predictors of radiographic severity and functional disability in Turkish patients with ankylosing spondylitis.

The objective of this study is to investigate the association between clinical and laboratory prognostic factors, radiographic severity, and functional limitations in Turkish patients with ankylosing spondylitis (AS). One hundred and two patients with AS were included in this study (66 male patients, 65%). All the necessary information regarding predictor variables, including clinical features, social status, and treatment regimens, were recorded diligently. Their spinal mobility was measured, and then, their disease activities were evaluated by using the Bath Ankylosing Spondylitis Disease Activity Index. Radiological damage (Bath Ankylosing Spondylitis Radiology Index, BASRI) and functional disability (Bath Ankylosing Spondylitis Functional Index, BASFI) were used to evaluate the outcome measures of AS. The male to female ratio was 1.8. Average age at symptom onset was 23.9 ± 28.24 years (6-54 years), and average disease duration was 16.15 ± 10.62 years. Occiput-to-wall distance, hand-to-floor distance, and the modified Schober's test results were worse in males. Hip involvement was more common in male patients, and all radiological measurements were worse in male patients than in the female ones. Disease duration, male sex, and renal stone occurrence were associated with higher radiological score. Erythrocyte sedimentation rate and the disease activity score were associated with high BASFI scores. Higher CRP levels and hip involvement were both associated with high BASRI and BASFI scores. Radiological hip involvement was determined to be an inauspicious predictor for AS (p < 0.0001). We determined a strong association of hip involvement, increased CRP levels, and renal stone history with severe radiographic damage. Hip involvement, disease duration, ESR, CRP levels, and lower socioeconomic status were all associated with a higher BASFI score.

Value of power Doppler and gray-scale US in the diagnosis of carpal tunnel syndrome: contribution of cross-sectional area just before the tunnel inlet as compared with the cross-sectional area at the tunnel.

OBJECTIVE: To determine the value of gray-scale and power Doppler ultrasonography in the evaluation of carpal tunnel syndrome (CTS). MATERIALS AND METHODS: Median nerves at the carpal tunnel were evaluated by using gray-scale and power Doppler ultrasonography and by using accepted and new criteria in 42 patients with CTS (62 wrists) confirmed by electromyogram and 33 control subjects. We evaluated the cross-sectional area of the nerve just proximal to the tunnel inlet (CSAa), and at mid level (CSAb). We then calculated the percentage area increase of CSAb, and area difference (CSAb-CSAa). We measured two dimensions of the nerve at the distal level to calculate the flattening ratio. The power Doppler ultrasonography was used to assess the number of vessels, which proceeded to give a score according to the vessel number, and lastly evaluated the statistical significance by comparing the means of patients with control subjects by the Student t test for independent samples. Sensitivities and specificities were determined for sonographic characteristics mentioned above. We obtained the receiver operating characteristic (ROC) curve to assess the optimal cut-off values for the diagnosis of CTS. RESULTS: A statistically significant difference was found between patients and the control group for mean CSAb, area difference, percentage area increase, and flattening ratio (p < 0.001, p < 0.001, p < 0.001, p < 0.05, respectively). From the ROC curve we obtained optimal cut-off values of 11 mm(2) for CSAb, 3.65 for area difference, 50% for the percentage of area increase, and 2.6 for the flattening ratio. The mean number of vessels obtained by power Doppler ultrasonography from the median nerve was 1.2. We could not detect vessels from healthy volunteers. Mean CSAbs related to vascularity intensity scores were as follows: score 0: 12.3 ± 2.8 mm(2), score 1: 12.3 ± 3.1 mm(2), score 2: 14.95 ± 3.5 mm(2), score 3: 19.3 ± 3.8 mm(2). The mean PI value in vessels of the median nerve was 4.1 ± 1. CONCLUSION: Gray-scale and power Doppler ultrasonography are useful in the evaluation of CTS.

Factors affecting swallowing during thyroid fine-needle aspiration biopsy.

OBJECTIVE: This prospective study was conducted to examine the factors that affect swallowing during thyroid fine-needle aspiration biopsy (FNAB). METHODS: Consecutive patients who were to undergo FNAB were approached for inclusion in the study. The depth and size of the nodule were measured in all patients. In a random fashion, patients were told not to swallow during the procedure, or they were not given any instructions regarding swallowing. RESULTS: Regarding the 143 participants in the study, no significant differences in age (P = .07), diameter and depth of the thyroid nodules (P = .14; P = .46, respectively), or cytologic diagnosis of thyroid aspirates (P > .20) were found between the swallowing (n = 50) and nonswallowing (n = 93) patient groups. Swallowing was observed in 29 patients who were instructed not to swallow (36%) and in 21 patients who were not given any instructions regarding swallowing (34%; P = .95). The mean duration of the procedure for swallowing patients was 19.8 seconds, and it was 15.7 seconds for nonswallowing patients. The duration of the procedure was significantly shorter in patients who did not swallow (P = .001). More male patients swallowed than female patients (P = .003). CONCLUSIONS: Giving specific instructions about not swallowing did not make a difference regarding swallowing by patients during the FNAB procedure. More male patients swallowed than female patients. A longer duration of the FNAB procedure meant that the operator was more likely to encounter swallowing (and thus displacement of the thyroid) while doing the procedure. Completing the procedure quickly is the surest method to avoid a moving thyroid while performing FNAB.

Fracture-dislocation of the hip: what does volumetric computed tomography add to detection, characterization, and planning treatment?

Traumatic hip fracture-dislocations are serious injuries and may lead to long-term disability even if managed correctly. Standard radiographic procedures have a limited value in fully evaluating a traumatized hip. Multidetector computed tomography is, today, a valuable diagnostic tool for the evaluation of patients with traumatic hip fracture-dislocation. This article reviews the mechanism, clinical presentation, diagnostic algorithm, imaging findings, and the treatment associated with hip fracture-dislocations. The role of volumetric computed tomography in detection, characterization, and planning treatment of these injuries is discussed.

Irreversiblity of remodeled features on high-resolution computerized tomography scans of asthmatic patients on conventional therapy: a 6-year longitudinal study.

Airway remodeling can be assessed using high-resolution computerized tomography (HRCT) scanning of both parenchymal-and airway abnormalities in patients with asthma. The aim of this study was to examine structural changes in large and small airways of asthmatic patients using HRCT to determine if remodeling changes had occurred after prolonged use of conventional anti-asthma therapy. HRCT scans were evaluated prospectively for evidence of the following abnormalities: bronchial wall thickening (BWT), bronchiectasis, mucoid impactions, small centrilobular opacities, thick linear opacities, focal hyperlucency, and emphysema. Fifty mild and moderate asthmatics were enrolled in the study group. These abnormalities were re-evaluated in the patients after the passage of 6 years of regular anti-asthma medication. Forty-six of the patients completed the study. The probability of finding at least one abnormality by HRCT investigation was statistically higher in the second scan than in the first (26 patients [56.5%] versus 18 patients [39.1%], p = 0.02]. Irreversibility ratios of abnormalities were 80%, 100%, 75%, 87.7%, 77.8%, and 100% for BWT, bronchiectasis, small centrilobular opacities, focal hyperlucency, thick linear opacity, and emphysema, respectively. The ratios for newly detected structural abnormalities were 25%, 2.5%, 0%, 7.9%, 8.1%, and 0% for BWT, bronchiectasis, small centrilobular opacities, focal hyperlucency, thick linear opacity, and emphysema, respectively. New occurrences and progression in BWT are associated with the duration of asthma affliction (p = 0.03). The results of our study indicate that HRCT remodeling features, once occurring, are irreversible in most of the patients, and new remodeling features also occur despite administering the standard asthma treatment.

Performance of the Wells and Revised Geneva scores for predicting pulmonary embolism.

OBJECTIVE: The objective of the study was to compare two scoring methods to predict the risk of pulmonary embolism (PE) as diagnosed with computed tomography angiography (CTA) and/or CT venography (CTV). METHODS: Prospectively over a 8-month period, emergency department patients and hospital inpatients with suspected PE were consecutively examined and their Wells and Revised Geneva scores calculated to stratify them into a risk group for PE probability. Then all patients were examined with CTA and CTV to determine the presence or absence of PE, as diagnosed by experienced radiology staff physicians. RESULTS: During the study period, 167 patients were suspected of having a PE and were interviewed for the calculation of their Wells and Revised Geneva scores. All patients underwent CTA or CTV, but the images of only 148 patients were adequate enough to make a certain diagnosis regarding PE. The data of these 148 patients were used for the study. The rates of PE in high, moderate, and low PE risk groups determined according to the Wells score and the Revised Geneva score were 89.6, 26.4, 7.8 and 83.3, 25.6, 0%, respectively. Among both inpatients and ED patients the area under the Wells score receiver operating characteristic curve was higher (P=0.04). When data from only ED patients were analyzed (104 patients) the scoring systems was not significantly different (P=0.07). CONCLUSION: The Wells rule seems to be more accurate among both inpatients and emergency department patients. The Revised Geneva score can be used in emergency department patients with high reliability.