THE EFFECTS OF REDUCED FORMAT AND TRAINING ON DETECTION OF SIMULATED LUNG NODULES IN X RAY IMAGES

Abstract

This dissertation addresses two problems dealing with detection sensitivity in radiology. First, is a 100-mm format adequate for displaying enhanced radiographic images in a photoelectronic imaging system and can the same format be used for permanent storage of these images? Second, what are the effects of training on the radiologist's sensitivity? Controlled 14-in. x 17-in. x-ray images were reduced to 100 mm to investigate which factors caused the decrement noted in the radiology literature. The 14-in. x 17-in. images used were generated from a 3M phantom. Three levels of kilovolt potential (80 kVp, 110 kVp, and 140 kVp) and two types of film (RPR and Cronex) were used to make the images. Twenty-four images, six for each of four quadrants, contained a simulated lung nodule and six contained no nodule. The 100-mm images were precisely controlled, miniaturized images of the above films. The task of the 16 non-radiologists was to decide if a simulated nodule was present or not. A signal detection experimental paradigm with a 10-point certainty scale was used for analysis. Correction for location was also implemented. For the first problem, results showed that kilovolt potential and film type interacted with the reduction procedure to change the direction of sensitivity (both in the positive and the negative directions) in the reduced format stimulus set, thereby indicating that the 100-mm format was not the main reason for reduced sensitivity noted in the literature. Our conclusion was that the 100-mm format would be suitable for image enhancement displays and for permanent hard copy in a photoelectronic radiologic imaging system. The second problem explores the effects of training on detection and sensitivity. The responses of the untrained observers to the 14-in. x 17-in. images were compared to previously gathered data of the staff and residents of the Arizona Health Sciences Center Radiology Department. Results indicated that the residents, even before formal training, were much better than the untrained observers, but well below the staff, in detection sensitivity. Within 6 months to a year later, they were equivalent to the staff. The main conclusion was that sensitivity does increase with training fairly rapidly and that the findings would be useful in designing further studies to investigate means to speed up or enhance the procedures used to train new radiologists.