What is the minimum amount of simulated breast movement required for visual detection of blurring? An exploratory investigation

Effect of simulated cataract on the accuracy of artificial intelligence in detecting diabetic retinopathy in color fundus photos

PURPOSE

Artificial intelligence (AI) is often trained on images without ocular co-morbidities, limiting its generalizability. This study aims to evaluate the accuracy of a convolutional neural network (CNN) applied to color fundus photos (CFPs) with simulated cataracts (SCs) in detecting diabetic retinopathy (DR).

METHODS

A database of 3662 CFPs (from Asia Pacific Tele-Ophthalmology Society (APTOS) 2019) was used. Using transfer learning, a CNN was trained to classify the training images as either DR or non-DR. The CNN was then applied to classify the testing images after an SC was applied, using varying degrees of Gaussian blur.

RESULTS

Accuracy without SC was 97.0%, sensitivity (Sn) 95.7%, specificity (Sp) 98.3%. For mild SC, accuracy was 93.1%, Sn 91.8%, Sp 94.3%. For moderate SC, accuracy was 62.8%, Sn 31.4%, Sp 95.2%. For severe SC, accuracy was 53.5%, Sn 11.8%, Sp 96.5%.

CONCLUSION

SCs significantly impaired AI accuracy. To prepare AI for clinical use, cataracts and other real-world clinical challenges affecting image quality must be accounted for.

Translating research evidence into clinical practice within a single breast imaging unit - A personal reflection

Health care organisations are required to deliver high quality evidence based and cost effective clinical care. Generating research evidence, appraising new published evidence and integrating it into clinical practice is time-consuming but an essential component in the continuous improvement of care delivery. This article is a personal reflection on the impact of research in clinical practice in a busy, single breast imaging unit from a number of different perspectives. It also highlights a number of benefits that are to be realised both from a service delivery perspective and across the whole staff team from engagement in such activity, which is an integral aspect of our work.

The impact of simulated motion blur on lesion detection performance in full-field digital mammography

OBJECTIVE

Motion blur is a known phenomenon in full-field digital mammography, but the impact on lesion detection is unknown. This is the first study to investigate detection performance with varying magnitudes of simulated motion blur.

METHODS

7 observers (15 ± 5 years' reporting experience) evaluated 248 cases (62 containing malignant masses, 62 containing malignant microcalcifications and 124 normal cases) for 3 conditions: no blurring (0 mm) and 2 magnitudes of simulated blurring (0.7 and 1.5 mm). Abnormal cases were biopsy proven. Mathematical simulation was used to provide a pixel shift in order to simulate motion blur. A free-response observer study was conducted to compare lesion detection performance for the three conditions. The equally weighted jackknife alternative free-response receiver operating characteristic was used as the figure of merit. Test alpha was set at 0.05 to control probability of Type I error.

RESULTS

The equally weighted jackknife alternative free-response receiver operating characteristic analysis found a statistically significant difference in lesion detection performance for both masses [F(2,22) = 6.01, p = 0.0084] and microcalcifications [F(2,49) = 23.14, p < 0.0001]. The figures of merit reduced as the magnitude of simulated blurring increased. Statistical differences were found between some of the pairs investigated for the detection of masses (0.0 vs 0.7 and 0.0 vs 1.5 mm) and all pairs for microcalcifications (0.0 vs 0.7, 0.0 vs 1.5 and 0.7 vs 1.5 mm). No difference was detected between 0.7 and 1.5 mm for masses.

CONCLUSION

The mathematical simulation of motion blur caused a statistically significant reduction in lesion detection performance. These false-negative decisions could have implications for clinical practice. Advances in knowledge: This research demonstrates for the first time that motion blur has a negative and statistically significant impact on lesion detection performance in digital mammography.

Blurred digital mammography images: an analysis of technical recall and observer detection performance

OBJECTIVE

Blurred images in full-field digital mammography are a problem in the UK Breast Screening Programme. Technical recalls may be due to blurring not being seen on lower resolution monitors used for review. This study assesses the visual detection of blurring on a 2.3-MP monitor and a 5-MP report grade monitor and proposes an observer standard for the visual detection of blurring on a 5-MP reporting grade monitor.

METHODS

28 observers assessed 120 images for blurring; 20 images had no blurring present, whereas 100 images had blurring imposed through mathematical simulation at 0.2, 0.4, 0.6, 0.8 and 1.0 mm levels of motion. Technical recall rate for both monitors and angular size at each level of motion were calculated. χ2 tests were used to test whether significant differences in blurring detection existed between 2.3- and 5-MP monitors.

RESULTS

The technical recall rate for 2.3- and 5-MP monitors are 20.3% and 9.1%, respectively. The angular size for 0.2- to 1-mm motion varied from 55 to 275 arc s. The minimum amount of motion for visual detection of blurring in this study is 0.4 mm. For 0.2-mm simulated motion, there was no significant difference [χ2 (1, N = 1095) = 1.61, p = 0.20] in blurring detection between the 2.3- and 5-MP monitors.

CONCLUSION

According to this study, monitors ≤2.3 MP are not suitable for technical review of full-field digital mammography images for the detection of blur. Advances in knowledge: This research proposes the first observer standard for the visual detection of blurring.

Analysis of motion during the breast clamping phase of mammography

OBJECTIVE

To measure paddle motion during the clamping phase of a breast phantom for a range of machine/paddle combinations.

METHODS

A deformable breast phantom was used to simulate a female breast. 12 mammography machines from three manufacturers with 22 flexible and 20 fixed paddles were evaluated. Vertical motion at the paddle was measured using two calibrated linear potentiometers. For each paddle, the motion in millimetres was recorded every 0.5 s for 40 s, while the phantom was compressed with 80 N. Independent t-tests were used to determine differences in paddle motion between flexible and fixed, small and large, GE Senographe Essential (General Electric Medical Systems, Milwaukee, WI) and Hologic Selenia Dimensions paddles (Hologic, Bedford, MA). Paddle tilt in the medial-lateral plane for each machine/paddle combination was calculated.

RESULTS

All machine/paddle combinations demonstrate highest levels of motion during the first 10 s of the clamping phase. The least motion is 0.17 ± 0.05 mm/10 s (n = 20) and the most motion is 0.51 ± 0.15 mm/10 s (n = 80). There is a statistical difference in paddle motion between fixed and flexible (p < 0.001), GE Senographe Essential and Hologic Selenia Dimensions paddles (p < 0.001). Paddle tilt in the medial-lateral plane is independent of time and varied from 0.04 ° to 0.69 °.

CONCLUSION

All machine/paddle combinations exhibited motion and tilting, and the extent varied with machine and paddle sizes and types.

ADVANCES IN KNOWLEDGE

This research suggests that image blurring will likely be clinically insignificant 4 s or more after the clamping phase commences.