Detection of small changes in medical and random-dot images comparing self-organizing map performance to human detection

Abstract

Radiologists use time-series of medical images to monitor the progression of a patient's conditions. They compare information gleaned from sequences of images to gain insight on progression or remission of the lesions, thus evaluating the progress of a patient's condition or response to therapy.Visual methods of determining differences between one series of images to another can be subjective or fail to detect very small differences. We propose the use of quantization errors obtained from selforganizing maps (SOM) for image content analysis. We tested this technique with MRI images to which we progressively added synthetic lesions. We have used a global approach that considers changes on the entire image as opposed to changes in segmented lesion regions only. We claim that this approach does not suffer from the limitations imposed by segmentation, which may compromise the results. Results show quantization errors increased with the increase in lesions on the images. The results are also consistent with previous studies using alternative approaches. We then compared the detectability ability of our method to that of human novice observers having to detect very small local differences in random-dot images. The quantization errors of the SOM outputs compared with correct positive rates, after subtraction of false positive rates ("guess rates"), increased noticeably and consistently with small increases in local dot size that were not detectable by humans. We conclude that our method detects very small changes in complex images and suggest that it could be implemented to assist human operators in image-based decision making.

Using the Quantization Error from Self-Organizing Map (SOM) Output for Fast Detection of Critical Variations in Image Time Series

“…In previous work (Wandeto et al, 2017a, Wandeto et al, 2017b, we have shown that the quantization error (QE) from SOM output after image learning can be effectively exploited for the fast and reliable detection of local changes in time series of medical images and satellite images for specific geographic regions of interest. The goal of the following proofof-concept study is to show that the QE varies consistently, reliably, and predictably with local variations in spatially distributed contrast signals in random-dot images and images with regularly distributed spatial contrast (geometric configuration).…”

Using Self Organizing Maps to Visualize Structural Change in Geographic Regions of Interest: Las Vegas across the Years

“…As shown in our previous work [4], the quantization error (QE) of the output of image analysis analyses of satellite image contents for the corresponding region of interest, can be provided in the light of consistent variations in the QE. SOM analysis on time series of satellite images is fast (less than two minutes for a series of images), and represents a promising technique for the automatic tracking and harvesting of landscape information in large bodies of satellite images.…”

“…Our previous work [4] had shown that a specific output variable of the SOM, the quantization error (QE), can be exploited as a diagnostic indicator for the presence of potentially critical local changes in medical image contents. In the present work, to further highlight the potential of this new computational approach, we used the QE output from SOM analyses of satellite images of Las Vegas, generated across the years 1984-2009.…”

“…This period is of particular interest as a major time of structural changes in the urban landscape of Las Vegas. As a reliable indicator of density, variability and local change in image contents [4], the QE is used here to reflect structural changes that took place in Las Vegas across these critical years: in the town as a whole, and in five of its major subdivisions. We also show how the QE output may be used further to meaningfully highlight demographic data for the same critical time period.…”

Occam’s Razor for Big Data? On Detecting Quality in Large Unstructured Datasets

“…A simple functional architecture of the self-organizing map may be applied to the unsupervised classification of massive amounts of patient data from different disease entities ranging from inflammation to cancer, as shown recently [35]. Other recent work [102][103][104][105] has shown the quantization error (QE) in the output of a basic self-organized neural network map (SOM); in short, the SOM-QE is a parsimonious and highly reliable measure of the smallest local change in contrast or color data in random-dot, medical, satellite, and other potentially large image data. The SOM is easily implemented, learns the pixel structure of any target image in about two seconds by unsupervised "winner-take-all" learning, and detects local changes in contrast or color in a series of subsequent input images with a to-the-single-pixel precision, in less than two seconds for a set of 20 images [106,107].…”