Paul R Goddard

Computed tomography in pulmonary emphysema

Fifty-three patients with chronic obstructive airways disease and 19 age-matched controls were studied using computed tomography (CT). The study shows that CT can detect the presence and distribution of pulmonary emphysema. Pulmonary vascular changes detectable on chest radiography correlate well with lung density as measured by CT. Patients with marked CT changes of emphysema had significantly greater impairment of diffusion capacity and FEV1.0/VC than the patients with less severe changes.

Diagnostic imaging of post-irradiation changes in the chest

Thirty-nine patients were studied with regard to post-irradiation changes in the chest. Twenty of these were reviewed retrospectively and 19 studied prospectively. All patients had chest radiographs and computed tomography (CT) of the chest following radiotherapy. Nineteen also had ventilation and perfusion studies of the lung, including single photon emission computed tomography (SPECT) and these were correlated with the chest radiographs and computed tomography. The majority showed abnormalities on computed tomography, the commonest being areas of lung opacification and evidence of volume loss. Several patients also showed a reduction in the size of pulmonary vessels. In most but not all, the changes were also seen on the chest radiographs. Abnormalities were not confined to the radiation fields, the vascular changes being present in large areas of lung which had not been directly irradiated. The structural and functional abnormalities correlated well as shown by ventilation and perfusion scintigraphy. However, single photon emission computed tomography was more sensitive than planar scintigraphy in showing perfusion defects, and it also showed some defects in areas of lung which appeared normal on computed tomography and the chest film. Computed and photon emission tomography were considerably more sensitive than chest radiography in showing the changes due to irradiation. The chest radiograph is clearly an insensitive indicator of post-irradiation change in the lung. Functional abnormalities are more profound and extensive than the chest film suggests, even when it is positive. There are clear implications for the planning of radiotherapy fields affecting the chest in patients who have good prospects of long-term survival. The maximum damage is related to irradiation of the hilum or mediastinum and this should be avoided wherever possible.

Chest Radiograph Interpretation by Medical Students

AIM To assess the ability of final year medical students to interpret conventional chest radiographs. MATERIALS AND METHODS Ten conventional chest radiographs were selected from a teaching hospital radiology department library that were good radiological examples of common conditions. All were conditions that a medical student should be expected to recognize by the end of their training. One normal radiograph was included. The radiographs were shown to 52 final year medical students who were asked to describe their findings. RESULTS The median score achieved was 12.5 out of 20 (range 6-18). There was no difference between the median scores of male and female students (12.5 and 12.3, respectively, p=0.82) but male students were more likely to be certain of their answers than female students (median certainty scores 23.0 and 14.0, respectively). The overall degree of certainty was low. On no radiograph were more than 25% of students definite about their answer. Students had received little formal radiology teaching (2-42 h, median 21) and few expressed an interest in radiology as a career. Only two (3.8%) students thought they were good at interpreting chest radiographs, 17 (32.7%) thought they were bad or awful. CONCLUSION Medical students reaching the end of their training do not perform well at interpreting simple chest radiographs. They lack confidence and have received little formal radiological tuition. Perhaps as a result, few are interested in radiology as a career, which is a matter for concern in view of the current shortage of radiologists in the UK.

CT appearances of relapse of lymphoma in the lung

The patterns of pulmonary relapse were studied in 15 patients with Hodgkins disease and one patient with Non-Hodgkins lymphoma. All the patients with Hodgkins disease had mediastinal lymphadenopathy at initial diagnosis. For those patients with no prior episodes of relapse the mean time to pulmonary involvement was 2 years 11 months compared to over 8 years for those who first relapsed elsewhere. Thirteen patients are still alive; five have been followed up for more than 2 years. Pulmonary involvement consisted of: 1. nodules, either central (12 patients) or peripheral (10 patients), often with connection to the pleura or mediastinum, and sometimes with cavitation; 2. consolidation with or without cavitation (four patients); 3. mediastinal extension into the lung parenchyma (three patients). In seven patients there was evidence of newly enlarged mediastinal nodes. Pleural effusions or masses were seen in six patients and pericardial involvement in one. The chest wall was involved in five. A combination of three or more of these signs were present in 11 patients. The pattern of involvement seen in the patient with Non-Hodgkins lymphoma was indistinguishable from that seen in Hodgkins disease. This study has demonstrated a variety of CT appearances useful in establishing or suggesting the diagnosis of pulmonary relapse. Enlarged mediastinal nodes were not a prerequisite but had been present in all patients at some stage in the course of the disease.

MRI of the cervical spine: Rheumatoid arthritis compared with cervical spondylosis

This study is a comparison of the cervical spine MR images from 26 patients with rheumatoid arthritis of the cervical spine with those from an age and sex matched group suffering from cervical spondylosis. Erosion of bone and major atlanto-axial subluxation were confined to rheumatoid arthritis. Soft tissue changes revealed by MRI included distortion of normal ligaments and bursae around the dens, particularly in rheumatoid arthritis. Abnormal masses of soft tissue were found in both groups, but those suggesting acute inflammation were much more frequent in rheumatoid arthritis than in cervical spondylosis. Neural compression was well demonstrated, and in rheumatoid arthritis was usually caused by bony structures whereas in cervical spondylosis it was usually due to disc material. It is concluded that MRI should be used as the first investigation to follow plain films in rheumatoid arthritis of the cervical spine. Bone and soft tissue changes are clearly shown, but interpretation of the images requires the recognition that some observed abnormalities may be due to coincidental cervical spondylosis.

The role of magnetic resonance imaging in cystic fibrosis

Magnetic resonance imaging (MRI) of the chest was performed on eight patients with known cystic fibrosis. Comparison was made with the findings on plain chest radiographs. MRI shows a greater extent of disease than that predicted by the chest films alone and clearly demonstrates peribronchial thickening and mucoid impacted bronchi. It also distinguishes between hilar lymphadenopathy and enlarged proximal pulmonary vessels as the cause of prominent hilar shadows without the need for contrast injection.

Fat-suppression magnetic resonance imaging in the preoperative localization of parathyroid adenomas.

Prospective localization of parathyroid adenomas was attempted in 16 patients with hyperparathyroidism prior to surgery. All patients had magnetic resonance imaging (MRI) using T1-weighted spin-echo (SE) sequences and a fat-suppression sequence, the short-tau inversion recovery (STIR) sequence. Correlation with the results of surgery yielded an overall sensitivity of 71% and a specificity of 94%. Performance was good in patients with previous surgery and with ectopic tumours. We believe that fat-suppression MRI is a valuable technique in the preoperative localization of parathyroid adenomas in patients with hyperparathyroidism.

Review of the use of MRI in soft tissue lesions

The MR images of 34 patients with soft tissue lesions were retrospectively evaluated to assess the accuracy of the technique in distinguishing benign from malignant lesions, and to assess the usefulness of various criteria in making this distinction. The overall sensitivity for the detection of malignancy was 75% with a specificity of 94%. Size of lesion was found to be a good criterion in predicting malignancy, lesion margin and signal intensity were less useful. The tissue type was determined in a few instances where signal characteristics were typical, notably lipomas and neural tumours, but this was not reliable and in most lesions the tissue of origin cannot be determined on MR imaging and biopsy is necessary.

Original PaperMRI of the cervical spine: Rheumatoid arthritis compared with cervical spondylosis

This study is a comparison of the cervical spine MR images from 26 patients with rheumatoid arthritis of the cervical spine with those from an age and sex matched group suffering from cervical spondylosis. Erosion of bone and major atlanto-axial subluxation were confined to rheumatoid arthritis. Soft tissue changes revealed by MRI included distortion of normal ligaments and bursae around the dens, particularly in rheumatoid arthritis. Abnormal masses of soft tissue were found in both groups, but those suggesting acute inflammation were much more frequent in rheumatoid arthritis than in cervical spondylosis. Neural compression was well demonstrated, and in rheumatoid arthritis was usually caused by bony structures whereas in cervical spondylosis it was usually due to disc material. It is concluded that MRI should be used as the first investigation to follow plain films in rheumatoid arthritis of the cervical spine. Bone and soft tissue changes are clearly shown, but interpretation of the images requires the recognition that some observed abnormalities may be due to coincidental cervical spondylosis.

Unsupervised learning in radiology using novel latent variable models

In this paper we compare a variety of unsupervised probabilistic models used to represent a data set consisting of textual and image information. We show that those based on latent Dirichlet allocation (LDA) out perform traditional mixture models in likelihood comparison. The data set is taken from radiology; a combination of medical images and consultants reports. The task of learning to classify individual tissue, or disease types, requires expert hand labeled data. This is both: expensive to produce and prone to inconsistencies in labeling. Here we present methods that require no hand labeling and also automatically discover sub-types of disease. The learnt models can be used for both prediction and classification of new unseen data.