Efficient Breast Cancer Diagnosis from Complex Mammographic Images Using Deep Convolutional Neural Network.

Medical image analysis places a significant focus on breast cancer, which poses a significant threat to women's health and contributes to many fatalities. An early and precise diagnosis of breast cancer through digital mammograms can significantly improve the accuracy of disease detection. Computer-aided diagnosis (CAD) systems must analyze the medical imagery and perform detection, segmentation, and classification processes to assist radiologists with accurately detecting breast lesions. However, early-stage mammography cancer detection is certainly difficult. The deep convolutional neural network has demonstrated exceptional results and is considered a highly effective tool in the field. This study proposes a computational framework for diagnosing breast cancer using a ResNet-50 convolutional neural network to classify mammogram images. To train and classify the INbreast dataset into benign or malignant categories, the framework utilizes transfer learning from the pretrained ResNet-50 CNN on ImageNet. The results revealed that the proposed framework achieved an outstanding classification accuracy of 93%, surpassing other models trained on the same dataset. This novel approach facilitates early diagnosis and classification of malignant and benign breast cancer, potentially saving lives and resources. These outcomes highlight that deep convolutional neural network algorithms can be trained to achieve highly accurate results in various mammograms, along with the capacity to enhance medical tools by reducing the error rate in screening mammograms.

Is Diagnostic Performance of Quantitative 2D-Shear Wave Elastography Optimal for Clinical Classification of Benign and Malignant Thyroid Nodules?: A Systematic Review and Meta-analysis.

RATIONALE AND OBJECTIVE: This study is a dedicated 2D-shear wave elastography (2D-SWE) review aimed at systematically eliciting up-to-date evidence of its clinical value in differential diagnosis of benign and malignant thyroid nodules. METHODS: PubMed, Web of Science, and Scopus databases were searched for studies assessing the diagnostic value of 2D-SWE for thyroid malignancy risk stratification published until December 2016. The retrieved titles and abstracts were screened and evaluated according to the predefined inclusion and exclusion criteria. Methodological quality of the studies was assessed using the Quality Assessment of Studies of Diagnostic Accuracy included in Systematic Review 2 (QUADAS-2) tool. Extracted 2D-SWE diagnostic performance data were meta-analyzed to assess the summary sensitivity, specificity, and area under the receiver operating characteristic curve. RESULTS: After stepwise review, 14 studies in which 2D-SWE was used to evaluate 2851 thyroid nodules (1092 malignant, 1759 benign) from 2139 patients were selected for the current study. Study quality on QUADAS-2 assessment was moderate to high. The summary sensitivity, specificity and area under the receiver operating characteristic curve of 2D-SWE for differential diagnosis of benign and malignant thyroid nodules were 0.66 (95% confidence interval [CI]: 0.64-0.69), 0.78 (CI: 0.76-0.80), and 0.851 (Q* = 0.85), respectively. The pooled diagnostic odds ratio, negative likelihood ratio, and positive likelihood ratio were 12.73 (CI: 8.80-18.43), 0.31 (CI: 0.22-0.44), and 3.87 (CI: 2.83-5.29), respectively. CONCLUSION: Diagnostic performance of quantitative 2D-SWE for malignancy risk stratification of thyroid nodules is suboptimal with mediocre sensitivity and specificity, contrary to earlier reports of excellence.

A margin-of-the-day online adaptive intensity-modulated radiotherapy strategy for cervical cancer provides superior treatment accuracy compared to clinically recommended margins: a dosimetric evaluation.

PURPOSE: To dosimetrically evaluate a margin-of-the-day (MoD) online adaptive intensity-modulated radiotherapy (IMRT) strategy for cervical cancer patients. The strategy is based on a single planning computed tomography (CT) scan and a pretreatment constructed IMRT plan library with incremental clinical target volumes (CTV)-to-planning target volumes (PTV) margins. MATERIAL AND METHODS: For 14 patients, 9-10 variable bladder filling CT scans acquired at pretreatment and after 40 Gy were available. Bladder volume variability during the treatment course was recorded by twice-weekly US bladder-volume measurements. A MoD strategy that selects the best IMRT plan of the day from a library of plans with incremental margins in steps of 5 mm was compared with a clinically recommended population-based margin (15 mm). To compare the strategies, for each fraction that had a recorded US bladder-volume measurement, the CT scan with the nearest bladder volume was selected from the pretreatment CT series and from the CT series acquired after 40 Gy. A frequency-weighted average of the dose-volume histograms (DVH) parameters calculated for the two selected CT scans was used to estimate the DVH parameters of the fraction of interest. RESULTS: The 15-mm recommended margin resulted in cervix-uterus underdosage in six of 14 patients. Compared with the 15-mm margin, the MoD strategy resulted in significantly better cervix-uterus coverage (p = 0.008) without a significant difference in the sparing of rectum, bladder, and small bowel. For each patient, 3-8 (median 5) plans were needed in the library of plans for the MoD strategy. The required range of the MoD was 5-45 mm (median 15 mm). Twenty-five percent of all fractions could be treated with a MoD of 5 mm and 81% of all fractions could be treated with a MoD up to 25 mm. CONCLUSIONS: Compared with a clinically recommended margin, a simple online adaptive strategy resulted in better cervix-uterus coverage without compromising organs at risk sparing.

Residual setup errors caused by rotation and non-rigid motion in prone-treated cervical cancer patients after online CBCT image-guidance.

PURPOSE: To quantify the impact of uncorrected or partially corrected pelvis rotation and spine bending on region-specific residual setup errors in prone-treated cervical cancer patients. METHODS AND MATERIALS: Fifteen patients received an in-room CBCT scan twice a week. CBCT scans were registered to the planning CT-scan using a pelvic clip box and considering both translations and rotations. For daily correction of the detected translational pelvis setup errors by couch shifts, residual setup errors were determined for L5, L4 and seven other points of interest (POIs). The same was done for a procedure with translational corrections and limited rotational correction (±3°) by a 6D positioning device. RESULTS: With translational correction only, residual setup errors were large especially for L5/L4 in AP direction (Σ=5.1/5.5mm). For the 7 POIs the residual setup errors ranged from 1.8 to 5.6mm (AP). Using the 6D positioning device, the errors were substantially smaller (for L5/L4 in AP direction Σ=2.7/2.2mm). Using this device, the percentage of fractions with a residual AP displacement for L4>5mm reduced from 47% to 9%. CONCLUSIONS: Setup variations caused by pelvis rotations are large and cannot be ignored in prone treatment of cervical cancer patients. Corrections with a 6D positioning device may considerably reduce resulting setup errors, but the residual setup errors should still be accounted for by appropriate CTV-to-PTV margins.

Toward an individualized target motion management for IMRT of cervical cancer based on model-predicted cervix-uterus shape and position.

BACKGROUND AND PURPOSE: To design and evaluate a 3D patient-specific model to predict the cervix-uterus shape and position. METHODS AND MATERIALS: For 13 patients lying in prone position, 10 variable bladder filling CT-scans were acquired, 5 at planning and 5 after 40Gy. The delineated cervix-uterus volumes in 2-5 pre-treatment CT-scans were used to generate patient-specific models that predict the cervix-uterus geometry by bladder volume. Model predictions were compared to delineations, excluding those used for model construction. The prediction error was quantified by the margin required around the predicted volumes to accommodate 95% of the delineated volume and by the predicted-to-delineated surface distance. RESULTS: The prediction margin was significantly smaller (average 50%) than the margin encompassing the cervix-uterus motion. The prediction margin could be decreased (from 7 to 5mm at planning and from 10 to 8mm after 40Gy) by increasing (from 2 to 5) the number of CT-scans used for the model construction. CONCLUSION: For most patients, even with a model based on only two CT-scans, the prediction error was well below the margin encompassing the cervix-uterus motion. The described approach could be used to create prior to treatment, an individualized treatment strategy.

Increasing treatment accuracy for cervical cancer patients using correlations between bladder-filling change and cervix-uterus displacements: proof of principle.

PURPOSE: To investigate application of pre-treatment established correlations between bladder-filling changes and cervix-uterus displacements in adaptive therapy. MATERIALS AND METHODS: Thirteen cervical cancer patients participated in this prospective study. Pre-treatment, and after delivery of 40 Gy, a full bladder CT-scan was acquired, followed by voiding the bladder and acquisition of 4 other 3D scans in a 1h period with a naturally filling bladder (variable bladder filling CT-scans, VBF-scans). For the pre-treatment VBF-scans, linear correlations between bladder volume change and displacements of the tip of the uterus (ToU) and the center of mass (CoM) of markers implanted in the fornices of the vagina relative to the full bladder planning scan were established. Prediction accuracy of these correlation models was assessed by comparison with actual displacements in CT-scans, both pre-treatment and after 40 Gy. Inter-fraction ToU and marker-CoM displacements were derived from the established correlations and twice-weekly performed in-room bladder volume measurements, using a 3D ultrasound scanner. RESULTS: Target displacement in VBF-scans ranged from up to 65 mm in a single direction to almost 0mm, depending on the patient. For pre-treatment VBF-scans, the linear correlation models predicted the mean 3D position change for the ToU of 26.1 mm±10.8 with a residual of only 2.2 mm±1.7. For the marker-CoM, the 8.4 mm±5.3 mean positioning error was predicted with a residual of 0.9 mm±0.7. After 40Gy, the mean ToU displacement was 26.8 mm±15.8, while prediction based on the pre-treatment correlation models yielded a mean residual error of 9.0 mm±3.7. Target positioning errors in the fractioned treatments were very large, especially for the ToU (-18.5mm±11.2 for systematic errors in SI-direction). CONCLUSIONS: Pre-treatment acquired VBF-scans may be used to substantially enhance treatment precision of cervical cancer patients. Application in adaptive therapy is promising and warrants further investigation. For highly conformal (IMRT) treatments, the use of a full bladder drinking protocol results in unacceptably large systematic set-up errors.

Inter-fraction bladder filling variations and time trends for cervical cancer patients assessed with a portable 3-dimensional ultrasound bladder scanner.

BACKGROUND AND PURPOSE: For cervical cancer patients, bladder filling variations may result in inadequate EBRT target coverage, unless large safety margins are used. For a group of patients who received full bladder instructions, inter-fraction variations and time trends in bladder volume were quantified, and a 3D ultrasound (US) scanner was tested for on-line bladder volume measurements. METHODS AND MATERIALS: For 24 patients, the bladder volume was measured with US at the time of the planning CT scan, and twice weekly during the course of RT. Comparisons of US with planning CT were used to assess the bladder scanner accuracy. Patients were treated in prone on a belly board, EPID images were acquired to correlate set-up errors with bladder filling variations. RESULTS: Measured US and CT bladder volumes were strongly correlated (R = 0.97, slope 1.1 +/- 0.1). The population mean bladder volume at planning of 378 +/- 209 ml (1 SD) reduced to 109 +/- 88 ml (1 SD) in week 6, a reduction by 71% (average reduction 46 ml/week), revealing a large inter-fraction time trend. Intra-patient variation in bladder volume during RT was 168 ml (1 SD) (range 70-266 ml). Rotation around the LR axis was significantly correlated with bladder volume changes. CONCLUSIONS: Despite a full bladder instruction, bladder volumes reduced dramatically during treatment, implying large time trends in target position of these patients. The portable US scanner provides a quick and reliable measurement of the bladder volume, which might assist future online treatment adaptation.