The flicker principle for improving detection of chest lesions.

Abstract

Many chest and other lesions are missed in practice but obvious in retrospect. The problem frequently arises not from lack of visibility, but rather from below-normal perception when visual search efficiency is reduced by fatigue, distraction, etc. Automatic screening of roentgenograms would be the ideal answer, but major problems are involved of criteria selection and the at present overwhelming amounts of irrelevant information presented along with desired signals. A basic simplifying approach is to compare old with new roentgenograms of a given patient, to highlight differences corresponding to tumor growth, organ displacement, and other signs. Subtraction and other direct methods are well established technics already in use. The flicker method enhances the value of such a comparison. In this, images of the two films are viewed alternately by the observer, using an appropriate repetition rate and blanking means. Any changes between examinations then tend to appear in motion, as in cine. Human perception is much greater for moving than for still objects, so the method facilitates detection of pathologic signs. After visiting astronomy laboratories to determine current flicker technology, we performed experiments to determine the potential sensitivity of the basic method in chest roentgenology. Opaque disks of from 1 through 12 mm in diameter, in 1 mm steps, were placed in order on one spot of a chest roentgenogram, and sequential slide films obtained. These latter films were then compared, using a very simple flicker apparatus consisting of two slide projectors and an improvised shutter run at an 80 per minute flicker rate; the images were held as closely “in register” as possible. The following increments were consistently detected by three observers: 1. Change in disk diameters ... 1 mm. 2. Change in disk position ... 1 mm. The required lesion radiographic contrast was then evaluated. Polystyrene balls of various diameters were taped on patients' backs, and serial chest roentgenograms obtained. This work, still in progress, indicates a chest nodule 3 mm in diameter is visible when you know its location, but not uniquely detectable by ordinary direct film examination. These experiments indicate the flicker method is promising but reveal many technical problems which must be solved before practical application is possible. They stem basically from the fact that all changes in the roentgenogram produce motion signals, necessitating the use of special hardware and procedures. All but the diagnostically important parts of the viewed images must be as identical as possible and suitably positioned (“in register”), since incidental flickers are essentially high- and low-frequency noise which detracts from recognition of the desired signals. The following variables require specific attention in chest radiography. A. Film exposure