GravityNet for end-to-end small lesion detection.

This paper introduces a novel one-stage end-to-end detector specifically designed to detect small lesions in medical images. Precise localization of small lesions presents challenges due to their appearance and the diverse contextual backgrounds in which they are found. To address this, our approach introduces a new type of pixel-based anchor that dynamically moves towards the targeted lesion for detection. We refer to this new architecture as GravityNet, and the novel anchors as gravity points since they appear to be "attracted" by the lesions. We conducted experiments on two well-established medical problems involving small lesions to evaluate the performance of the proposed approach: microcalcifications detection in digital mammograms and microaneurysms detection in digital fundus images. Our method demonstrates promising results in effectively detecting small lesions in these medical imaging tasks.

Phantom study and clinical application of total-body 18F-FDG PET/CT imaging: How to use small voxel imaging better?

BACKGROUND: Conventional PET/CT imaging reconstruction is typically performed using voxel size of 3.0-4.0 mm in three axes. It is hypothesized that a smaller voxel sizes could improve the accuracy of small lesion detection. This study aims to explore the advantages and conditions of small voxel imaging on clinical application. METHODS: Both NEMA IQ phantom and 30 patients with an injected dose of 3.7 MBq/kg were scanned using a total-body PET/CT (uEXPLORER). Images were reconstructed using matrices of 192 × 192, 512 × 512, and 1024 × 1024 with scanning duration of 3 min, 5 min, 8 min, and 10 min, respectively. RESULTS: In the phantom study, the contrast recovery coefficient reached the maximum in matrix group of 512 × 512, and background variability increased as voxel size decreased. In the clinical study, SUVmax, SD, and TLR increased, while SNR decreased as the voxel size decreased. When the scanning duration increased, SNR increased, while SUVmax, SD, and TLR decreased. The SUVmean was more reluctant to the changes in imaging matrix and scanning duration. The mean subjective scores for all 512 × 512 groups and 1024 × 1024 groups (scanning duration ≥ 8 min) were over three points. One false-positive lesion was found in groups of 512 × 512 with scanning duration of 3 min, 1024 × 1024 with 3 min and 5 min, respectively. Meanwhile, the false-negative lesions found in group of 192 × 192 with duration of 3 min and 5 min, 512 × 512 with 3 min and 1024 × 1024 with 3 min and 5 min were 5, 4, 1, 4, and 1, respectively. The reconstruction time and storage space occupation were significantly increased as the imaging matrix increased. CONCLUSIONS: PET/CT imaging with smaller voxel can improve SUVmax and TLR of lesions, which is advantageous for the diagnosis of small or hypometabolic lesions if with sufficient counts. With an 18F-FDG injection dose of 3.7 MBq/kg, uEXPLORER PET/CT imaging using matrix of 512 × 512 with 5 min or 1024 × 1024 with 8 min can meet the image requirements for clinical use.

RenseNet: A Deep Learning Network Incorporating Residual and Dense Blocks with Edge Conservative Module to Improve Small-Lesion Classification and Model Interpretation.

Deep learning has become an essential tool in medical image analysis owing to its remarkable performance. Target classification and model interpretability are key applications of deep learning in medical image analysis, and hence many deep learning-based algorithms have emerged. Many existing deep learning-based algorithms include pooling operations, which are a type of subsampling used to enlarge the receptive field. However, pooling operations degrade the image details in terms of signal processing theory, which is significantly sensitive to small objects in an image. Therefore, in this study, we designed a Rense block and edge conservative module to effectively manipulate previous feature information in the feed-forward learning process. Specifically, a Rense block, an optimal design that incorporates skip connections of residual and dense blocks, was demonstrated through mathematical analysis. Furthermore, we avoid blurring of the features in the pooling operation through a compensation path in the edge conservative module. Two independent CT datasets of kidney stones and lung tumors, in which small lesions are often included in the images, were used to verify the proposed RenseNet. The results of the classification and explanation heatmaps show that the proposed RenseNet provides the best inference and interpretation compared to current state-of-the-art methods. The proposed RenseNet can significantly contribute to efficient diagnosis and treatment because it is effective for small lesions that might be misclassified or misinterpreted.

Advances in Breast PET Instrumentation.

Dedicated breast PET scanners currently have a spatial resolution in the 1.5 to 2 mm range, and the ability to provide tomographic images and quantitative data. They are also commercially available from a few vendors. A review of past and recent advances in the development and performance of dedicated breast PET scanners is summarized.

Improving Small Lesion Segmentation in CT Scans using Intensity Distribution Supervision: Application to Small Bowel Carcinoid Tumor.

Finding small lesions is very challenging due to lack of noticeable features, severe class imbalance, as well as the size itself. One approach to improve small lesion segmentation is to reduce the region of interest and inspect it at a higher sensitivity rather than performing it for the entire region. It is usually implemented as sequential or joint segmentation of organ and lesion, which requires additional supervision on organ segmentation. Instead, we propose to utilize an intensity distribution of a target lesion at no additional labeling cost to effectively separate regions where the lesions are possibly located from the background. It is incorporated into network training as an auxiliary task. We applied the proposed method to segmentation of small bowel carcinoid tumors in CT scans. We observed improvements for all metrics (33.5% → 38.2%, 41.3% → 47.8%, 30.0% → 35.9% for the global, per case, and per tumor Dice scores, respectively.) compared to the baseline method, which proves the validity of our idea. Our method can be one option for explicitly incorporating intensity distribution information of a target in network training.

Small lesion depiction and quantification accuracy of oncological 18F-FDG PET/CT with small voxel and Bayesian penalized likelihood reconstruction.

BACKGROUND: To investigate the influence of small voxel Bayesian penalized likelihood (SVB) reconstruction on small lesion detection compared to ordered subset expectation maximization (OSEM) reconstruction using a clinical trials network (CTN) chest phantom and the patients with 18F-FDG-avid small lung tumors, and determine the optimal penalty factor for the lesion depiction and quantification. METHODS: The CTN phantom was filled with 18F solution with a sphere-to-background ratio of 3.81:1. Twenty-four patients with 18F-FDG-avid lung lesions (diameter < 2 cm) were enrolled. Six groups of PET images were reconstructed: routine voxel OSEM (RVOSEM), small voxel OSEM (SVOSEM), and SVB reconstructions with four penalty factors: 0.6, 0.8, 0.9, and 1.0 (SVB0.6, SVB0.8, SVB0.9, and SVB1.0). The routine and small voxel sizes are 4 × 4 × 4 and 2 × 2 × 2 mm3. The recovery coefficient (RC) was calculated by dividing the measured activity by the injected activity of the hot spheres in the phantom study. The SUVmax, target-to-liver ratio (TLR), contrast-to-noise ratio (CNR), the volume of the lesions, and the image noise of the liver were measured and calculated in the patient study. Visual image quality of the patient image was scored by two radiologists using a 5-point scale. RESULTS: In the phantom study, SVB0.6, SVB0.8, and SVB0.9 achieved higher RCs than SVOSEM. The RC was higher in SVOSEM than RVOSEM and SVB1.0. In the patient study, the SUVmax, TLR, and visual image quality scores of SVB0.6 to SVB0.9 were higher than those of RVOSEM, while the image noise of SVB0.8 to SVB1.0 was equivalent to or lower than that of RVOSEM. All SVB groups had higher CNRs than RVOSEM, but there was no difference between RVOSEM and SVOSEM. The lesion volumes derived from SVB0.6 to SVB0.9 were accurate, but over-estimated by RVOSEM, SVOSEM, and SVB1.0, using the CT measurement as the standard reference. CONCLUSIONS: The SVB reconstruction improved lesion contrast, TLR, CNR, and volumetric quantification accuracy for small lesions compared to RVOSEM reconstruction without image noise degradation or the need of longer emission time. A penalty factor of 0.8-0.9 was optimal for SVB reconstruction for the small tumor detection with 18F-FDG PET/CT.

Evaluating deep learning methods in detecting and segmenting different sizes of brain metastases on 3D post-contrast T1-weighted images.

Purpose: We investigate the impact of various deep-learning-based methods for detecting and segmenting metastases with different lesion volume sizes on 3D brain MR images. Approach: A 2.5D U-Net and a 3D U-Net were selected. We also evaluated weak learner fusion of the prediction features generated by the 2.5D and the 3D networks. A 3D fully convolutional one-stage (FCOS) detector was selected as a representative of bounding-box regression-based detection methods. A total of 422 3D post-contrast T1-weighted scans from patients with brain metastases were used. Performances were analyzed based on lesion volume, total metastatic volume per patient, and number of lesions per patient. Results: The performance of detection of the 2.5D and 3D U-Net methods had recall of > 0.83 and precision of > 0.44 for lesion volume > 0.3 cm 3 but deteriorated as metastasis size decreased below 0.3 cm 3 to 0.58 to 0.74 in recall and 0.16 to 0.25 in precision. Compared the two U-Nets for detection capability, high precision was achieved by the 2.5D network, but high recall was achieved by the 3D network for all lesion sizes. The weak learner fusion achieved a balanced performance between the 2.5D and 3D U-Nets; particularly, it increased precision to 0.83 for lesion volumes of 0.1 to 0.3 cm 3 but decreased recall to 0.59. The 3D FCOS detector did not outperform the U-Net methods in detecting either the small or large metastases presumably because of the limited data size. Conclusions: Our study provides the performances of four deep learning methods in relationship to lesion size, total metastasis volume, and number of lesions per patient, providing insight into further development of the deep learning networks.

Physical characteristics of 68Ga DOTATATE PET/CT affecting small lesion detectability.

BACKGROUND AND PURPOSE: 68Ga DOTATATE PET/CT protocols are similar to 18F FDG protocols despite differences in physical properties, biodistribution, and tumor uptake. The purpose of this study is to evaluate the impact of scan time (counts), and target activity on signal-to-noise ratio (SNR) in various sized targets, or lesions. To evaluate this, phantom experiments and analysis of clinical 68Ga DOTATATE PET/CT studies were performed. MATERIALS AND METHODS: 68Ga was first compared to 18F in phantom studies to evaluate recovery coefficients and SNR. 68Ga phantom studies were also acquired in list mode, and at varying target activities to evaluate the effects of acquisition time and high target concentrations on SNR in clinically relevant small (8 mm) and larger targets (≥ 12 mm). Clinical studies (n = 50) were analyzed to determine if phantom target concentrations and SNR are present in clinical 68Ga DOTATATE studies at similarly very high tumor activity concentrations (n = 159). RESULTS: In phantoms, recovery coefficient and SUVmax for 68Ga were ~87% of 18F. SNR for 68Ga was ~65% of 18F. For the 68Ga small target (8 mm) at standard T/B = 2.4, increasing scan time from 5 to 15 minutes increased SNR from < 1 to 1.6, and did not result in target identification. Increasing T/B from 2.4 to 10.9, however, dramatically increased SNR from < 1 to 22.3. Increased T/B resulted in clear visibility of the 8 mm target, even for 1-minute scans. In patients, high hepatic tumor SUVmax (27.3±29.6), resulted in high SNR (12.5±9.8). For extrahepatic tumors, high SUVmax (41.6±42.8), resulted in high SNR (43.8±49.9). CONCLUSION: Very high target or T/B, even in small targets, can offset the physical limitations of 68Ga. High target uptake and high T/B are primary factors influencing small lesion detectability.

Contrast-Enhanced Spectral Mammography-Based Radiomics Nomogram for Identifying Benign and Malignant Breast Lesions of Sub-1 cm.

OBJECTIVES: To develop a radiomics nomogram that incorporates contrast-enhanced spectral mammography (CESM)-based radiomics features and clinico-radiological variables for identifying benign and malignant breast lesions of sub-1 cm. METHODS: This retrospective study included 139 patients with the diameter of sub-1 cm on cranial caudal (CC) position of recombined images. Radiomics features were extracted from low-energy and recombined images on CC position. The variance threshold, analysis of variance (ANOVA) and least absolute shrinkage and selection operator (LASSO) algorithms were used to select optimal predictive features. Radiomics signature (Rad-score) was calculated by a linear combination of selected features. The independent predictive factors were identified by ANOVA and multivariate logistic regression. A radiomics nomogram was developed to predict the malignant probability of lesions. The performance and clinical utility of the nomogram was evaluated by receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). RESULTS: Nineteen radiomics features were selected to calculate Rad-score. Breast imaging reporting and data system (BI-RADS) category and age were identified as predictive factors. The radiomics nomogram combined with Rad-score, BI-RADS category, and age showed better performance (area under curves [AUC]: 0.940, 95% confidence interval [CI]: 0.804-0.992) than Rad-score (AUC: 0.868, 95% CI: 0.711-0.958) and clinico-radiological model (AUC: 0.864, 95% CI: 0.706-0.956) in the validation cohort. The calibration curve and DCA showed that the radiomics nomogram had good consistency and clinical utility. CONCLUSIONS: The radiomics nomogram incorporated with CESM-based radiomics features, BI-RADS category and age could identify benign and malignant breast lesions of sub-1 cm.

Variable tilt-angle, parallel-hole collimation system for high-resolution molecular imaging gamma tomosynthesis.

PURPOSE: This study investigates a novel gamma tomosynthesis (GT) method based on a variable tilt-angle, parallel-hole collimator (VAPHC) which, mounting to a conventional gamma, is able to perform high-resolution three-dimensional imaging. METHODS: The VAPHC has the remarkable feature to be modular, consisting of independent collimation elements able to tilt according to variable angles [-45° to +45°]. Spatial resolutions were measured in reconstructed GT images using a point source at different source-to-collimator distances, while sensitivity was evaluated over the range of slant angles using a disk-source. Image contrast (IC) and contrast-to-noise-ratio (CNR) of sub-centimeters tumors were evaluated using a breast phantom containing a background activity and spheres filled with 99mTc to simulate lesions at two depths. Breast phantom GT images were compared with planar and circular-orbit SPECT acquisitions of equal scan-time. RESULTS: Planar spatial resolutions range from 9 to 14 mm over a depth range of 6-10 cm; spatial resolution in depth dimension becomes two times greater than those in the other dimensions. The measured sensitivity decreases from 9 cps/µCi to 6 cps/µCi varying the slant angle from 5° to 45°. The measured IC and CNR of GT reconstructed images demonstrated that it was possible to improve the spatial resolution/sensitivity trade-off. CONCLUSIONS: The proposed GT based VAPHC demonstrated the potential for superior spatial resolution and contrast compared to planar and SPECT acquisitions. A conventional gamma camera equipped with the VAPHC could be located at the minimum distance from the patient, thus improving detection, localisation and characterisation of sub-centimetre lesions.

Clinical evaluation of whole-body oncologic PET with time-of-flight and point-spread function for the hybrid PET/MR system.

PURPOSE: Hybrid positron emission tomography/magnetic resonance (PET/MR) imaging is a new multimodality imaging technology that can provide structural and functional information simultaneously. The aim of this study was to investigate the effects of the time-of-flight (TOF) and point-spread function (PSF) on small lesions observed in PET/MR images from clinical patient image sets. MATERIALS AND METHODS: This study evaluated 54 small lesions in 14 patients who had undergone 18F-fluorodeoxyglucose (FDG) PET/MR. Lesions up to 30mm in diameter were included. The PET data were reconstructed with a baseline ordered-subsets expectation-maximization (OSEM) algorithm, OSEM+PSF, OSEM+TOF and OSEM+TOF+PSF. PET image quality and small lesions were visually evaluated and scored by a 3-point scale. A quantitative analysis was then performed using the mean and maximum standardized uptake value (SUV) of the small lesions (SUVmean and SUVmax). The lesions were divided into two groups according to the long-axis diameter and the location respectively and evaluated with each reconstruction algorithm. We also evaluated the background signal by analyzing the SUVliver. RESULTS: OSEM+TOF+PSF provided the highest value and OSEM+TOF or PSF showed a higher value than OSEM for the visual assessment and quantitative analysis. The combination of TOF and PSF increased the SUVmean by 26.6% and the SUVmax by 30.0%. The SUVliverwas not influenced by PSF or TOF. For the OSEM+TOF+PSF model, the change in SUVmean and SUVmax for lesions <10mm in diameter was 31.9% and 35.8%, and 24.5% and 27.6% for lesions 10-30mm in diameter, respectively. The abdominal lesions obtained the higher SUV than those of chest on the images with TOF and/or PSF. CONCLUSION: Application of TOF and PSF significantly increased the SUV of small lesions in hybrid PET/MR images, potentially improving small lesion detectability.

Comparison of the effect between pegaptanib and ranibizumab on exudative age-related macular degeneration with small lesion size.

PURPOSE: To compare the effect of pegaptanib versus ranibizumab on exudative age-related macular degeneration (AMD) with small lesion size. METHODS: This is a retrospective study of 81 eyes from 78 patients with exudative AMD treated and followed up over 12 months. Patients with baseline best corrected visual acuity (BCVA) under 20/400 and with a greatest linear dimension of lesion over 4500 µm were excluded from the study. Twenty-six eyes from 25 patients were treated with three consecutive intravitreal injections of pegaptanib (IVP group) and 55 eyes from 54 patients were treated with three consecutive ranibizumab injections (IVR group). Each therapy was repeated as needed. The alteration in BCVA was evaluated in the IVP and IVR groups. RESULTS: No differences were detected in baseline parameters between the IVP and IVR groups. The mean BCVA (logMAR) at month 1, 3, 6 and 12 after the initial treatment was improved from baseline in the IVP group (-0.095, -0.17, -0.18 and -0.18, respectively) and in the IVR group (-0.077, -0.15, -0.17 and -0.11, respectively), which was statistically significant. There was no difference in the change in mean BCVA between IVP and IVR groups at the same time periods. CONCLUSIONS: The visual outcome of IVP was equivalent with IVR in exudative AMD with small lesion size.

Application of Mammotome minimally invasive biopsy in diagnosis of non-palpable small lesions in breast / 上海第二医科大学学报

Objective To evaluate the application of Mammotome minimally invasive biopsy(MMIBS) in the diagnosis of non-palpable small lesions in breast. Methods A total of 42 patients with non-palpable small breast lesions and abnormal imaging findings were performed MMIBS.After MMIBS,21 patients received open operations,and the pathological diagnosis was compared and analysed. Results All these 42 patients were successfully performed MMIBS,with no severe complications,and the pathological diagnosis of MMIBS were breast cancer(8 patients) and benign pathological changes(34 patients).Among the 21 patients with open operations,the post-operative pathological findings of 20 patients were same as the results of MMIBS,with the sensitivity of 88.9% and specificity of 100%.The other 21 patients without operations were followed up for 6 to 12 months,and no abnormal mammographic and ultrasonographic findings were observed.Conclusion MMIBS boasts the characteristics of minimal invasion,safety,high sensitivity and high specificity,which can serve as the first choice in the diagnosis of non-palpable small lesions with abnormal imaging findings.

Endoscopic and clinicopathologic analysis of small advanced colorectal carcinomas less than 10 nun in maximum diameter / 中华消化内镜杂志

Objective Twelve cases with small advanced cole-rectal cancer less than 10 mm in diameter were examined. Methods The endoscopic findings and clinicopathologic data of 12 cases with advanced colorectal cancer less than 10 mm in diameter were compared with those larger than 10 mm in diameter. Results Approximated estimation of the macroscopic types were Ⅱ c in 5 cases, Ⅱ a in 3 cases, and I s in 4 cases respectively. As for the distribution of these lesions, there were 5 in sigmoid colon, 4 in transverse colon, 1 in cecum, and 2 in descending colon. No lesions were detected in rectum. Histologic grade at the deepest invasive portion was shown well, moderately, and poorly differentiated in 2, 8 and 1 case respectively and another case with mucinous adenocarcinomas. Lymphnode metastasis was present in 4 of 12 lesions (33% ). This incidence is rather high, as regards to the small size of each lesion. Endoscopi-cally, all lesions accompanied with converging folds and 7 lesions examined by magnifying colonoscope showed V N pit pattern. Conclusion These results indicated that the advanced colorectal cancers less than 10mm in diameter showing similar characteristics with those of superficial depressed type cancer with likely is the precurser of advanced cancer. The findings of converging folds and pit pattern are considered to be the useful indicators for estimating the depth 'of invasion.

Microlarngoscopic Operation (with 62 cases report) / 昆明医科大学学报

This artical analyses 62cases of microlarngoscopic operation under general anesthesia. We preliminarly conclude that this method has many advantages: better visual operation field,meticulous technic, excellent voice restored after operation. By means of this method, very small lesion that can' t be discovered by indirect laryngoscope or even fibrolaryngoscope could be discovered and treated in the early stage.