O'Connor, Michael K

AbstractObjectiveWe evaluated the accuracy of molecular breast imaging (MBI)—a nuclear medicine technique that employs dedicated dual-detector, cadmium zinc telluride gamma cameras to image the functional uptake of a radiopharmaceutical (typically Tc-99m sestamibi) in the breast—in patients with suspicious calcifications on mammography.MethodsWomen scheduled for stereotactic biopsy of calcifications detected on 2D digital mammography were prospectively enrolled to undergo MBI before biopsy. Molecular breast imaging was performed with injection of Tc-99m sestamibi and a dual-detector, cadmium zinc telluride gamma camera. Positive findings on either modality were biopsied. High-risk and malignant biopsy findings were excised.ResultsIn 71 participants, 76 areas of calcifications were recommended for biopsy after mammography, and 24 (32%) were malignant, including 20 cases of ductal carcinoma in situ (DCIS) and 4 cases of invasive ductal cancer. Prebiopsy MBI was positive in 17 of the 76 (22%) calcifications, including 10 of 20 (50%) DCISs and 2 of 4 (50%) invasive cancers. The median pathologic size for MBI–positive cancers was 1.5 cm (range 0.5–3.2 cm) compared with 0.9 cm (range 0.1–2.0 cm) for MBI–negative cancers (P = 0.09).Non-mass uptake on MBI led to additional biopsies of 6 sites in 6 patients, and 2 of 6 (33%) MBI–detected incidental lesions showed malignancy; both DCIS contralateral to the mammographically detected calcifications. The overall per-lesion positive and negative predictive values of MBI in this prebiopsy setting were 61% (14 of 23) and 80% (47 of 59), respectively.ConclusionMolecular breast imaging has insufficient negative predictive value to identify calcifications in which biopsy could be avoided. However, among women presenting for biopsy of suspicious calcifications, MBI revealed additional sites of mammographically occult breast cancer.To avoid biopsy of suspicious calcifications on mammography, negative findings on MBI should not be used.

PURPOSE Background parenchymal uptake (BPU), which describes the level of radiotracer uptake in normal fibroglandular tissue on molecular breast imaging (MBI), has been identified as a breast cancer risk factor. Our objective was to develop and validate a deep learning model using image convolution to automatically categorize BPU on MBI. METHODS MBI examinations obtained for clinical and research purposes from 2004 to 2015 were reviewed to classify the BPU pattern using a standardized five-category scale. Two expert radiologists provided interpretations that were used as the reference standard for modeling. The modeling consisted of training and validating a convolutional neural network to predict BPU. Model performance was summarized in data reserved to test the performance of the algorithm at the per-image and per-breast levels. RESULTS Training was performed on 24,639 images from 3,133 unique patients. The model performance on the withheld testing data (6,172 images; 786 patients) was evaluated. Using direct matching on the predicted classification resulted in an accuracy of 69.4% (95% CI, 67.4% to 71.3%), and if prediction within one category was considered, accuracy increased to 96.0% (95% CI, 95.2% to 96.7%). When considering the breast-level prediction of BPU, the accuracy remained strong, with 70.3% (95% CI, 68.0% to 72.6%) and 96.2% (95% CI, 95.3% to 97.2%) for the direct match and allowance for one category, respectively. CONCLUSION BPU provided a robust target for training a convolutional neural network. A validated computer algorithm will allow for objective, reproducible encoding of BPU to foster its integration into risk-stratification algorithms.

ABSTRACT BACKGROUND AND PURPOSE SISCOM and STATISCOM were clinically proved to be effective for ictal/inter‐ictal single‐photon emission computed tomography (SPECT) analysis coregistered with magnetic resonance imaging (MRI) for seizure localization. Recently, a software package also became available for this analysis. This study aimed to investigate and compare the performance of these analysis methods for seizure localization. METHODS A total of 378 patients who underwent 99m Tc‐ethyl cysteinate dimer (ECD) SPECT scans were retrospectively reviewed and 28 remained after applying exclusion criteria. Their SPECT and MRI images were analyzed with SISCOM (with z ‐score of 1.5 and 2), STATISCOM, and MIMneuro, resulting in a total of 112 image data sets. Two experienced radiologists participated in the blind review process using a custom tool and they can mark up to two hyper‐ and/or hypoperfusion regions. Their review results were analyzed using the Jackknife Free Response Receiver‐Operating Characteristics (JAFROC) test and the JAFROC figure‐of‐merit (FoM) was reported for each method. The interobserver agreement was also assessed using Cohen's kappa test. RESULTS Based on the readers’ two choices, averaged FoM was 85.7%, 83.9%, 66.1%, and 51.8% for STATISCOM, MIMneuro, SISCOM ( z ‐score = 2), and SISCOM ( z ‐score = 1.5), respectively. The average confidence rating was 2.5, 2.3, 1.6, and 1.1 for STATISCOM, MIMneuro, SISCOM ( z ‐score = 2), and SISCOM ( z ‐score = 1.5), respectively. For interobserver agreement, kappa was .742 for STATISCOM, .816 for MIMneuro, .517 for SISCOM ( z ‐score = 2), and .441 for SISCOM ( z ‐score = 1.5; all P < .001). CONCLUSION Our study demonstrated that STATISCOM showed the best performance for seizure localization, which was closely followed by MIMneuro. In addition, MIMneuro was not inferior to SISCOM with either z ‐score.

Transplantation of gene-corrected autologous hepatocytes can cure metabolic disease in a preclinical pig model of hereditary tyrosinemia type 1.