Optimization of stereotactic radiotherapy (SRT) planning for brain metastases from lung cancer by Cyberknife system: reducing dose distribution inner ear.

Objective: Through retrospective statistical analysis of radiation distribution in inner ear avoidance for brain metastases from lung cancer by the CyberKnife (CK) system, it can provide a reference for stereotactic radiotherapy (SRT) planning and treatment optimization. Methods: Computed tomography/magnetic resonance imaging data of 44 patients with one brain metastases lesion from lung cancer were used to re-plan and analyze, who had been treated by CK system from April 2021 to April 2022. The prescribed doses of 14-30 Gy in 1-3 fractions was simultaneously delivered to the metastatic lesions. The SRT plans for the same patients were replaned under with and without inner ear avoidance setting. The plan parameters and dose distribution differences were compared between plans. Results: All plans met the dose restrictions. There were no significant differences in the coverage (Coverage), conformity index (CI), mean dose (Dmean), the maximum dose (Dmax) and minimum dose (Dmin) of planning target volume (PTV). With inner ear avoidance setting, the Dmax and Dmean of inner ear area decreased by 13.76% and 12.15% (p<0.01), respectively. The total number of machine nodes and monitor units (MU) increased by 4.63% and 1.06%. Conclusions: During the SRT plan designing for brain metastases from lung cancer, the dose distribution in inner ear area could be reduced by avoidance setting, and the patient's hearing would be well protected.

Accuracy analysis of different dose calculation algorithms for locally advanced pancreatic cancer stereotactic body radiotherapy.

Background: The dose distribution in different optimization algorithm plans of stereotactic radiotherapy (SBRT) for locally advanced pancreatic cancer (LAPC) were compared and analyzed using monte carlo dose calculate algorithm (MC). Methods: A retrospective study analyzed 26 LAPC patients treated with SBRT. The SBRT plans were designed by raytracing (RT) and fine size pencil beam (FSPB) algorithms in the CyberKnife (CK) precision system, all of which met the requirements of clinical target dose and organ at risk (OAR). Keeping the original optimization parameters unchanged, the RT and FSPB algorithm plans were recalculated by MC algorithm. The accuracy of different algorithm plnas were compared and analyzed by using planning parameters and dose distribution. Results: There was no significant differences in the coverage and conformal index (CI) of the planned target volume (PTV) between RT and FSPB algorithm plans, but dose distribution of organ at risk (OAR) and the maximum dose outside the PTV boundary of 2 cm (D2cm) were lower in FSPB plans compared to RT plans, and this difference was statistically significant with p-values < 0.05. Compared to the MC algorithm, both RT algorithm and FSPB algorithm overestimated dose of the PTV and OAR. The RT algorithm was more consistent with the MC algorithm than the FSPB algorithm. The relative error of PTV coverage within the RT algorithm was 8.02% ± 1.53%, and the relative error range of OAR dose parameters was 3.32% -12.73%. Conclusion: Although the FSPB algorithm could achieve rapid dose drop-off around the PTV and lower dose distribution in the OAR for pancreatic cancer SBRT plans, the algorithm error were higher than the RT algorithm. RT and FSPB algorithm overestimated the dose in the target and OAR. That was important to evaluate the clinical plans.

A deep learning approach for automatic tumor delineation in stereotactic radiotherapy for non-small cell lung cancer using diagnostic PET-CT and planning CT.

Introduction: Accurate delineation of tumor targets is crucial for stereotactic body radiation therapy (SBRT) for non-small cell lung cancer (NSCLC). This study aims to develop a deep learning-based segmentation approach to accurately and efficiently delineate NSCLC targets using diagnostic PET-CT and SBRT planning CT (pCT). Methods: The diagnostic PET was registered to pCT using the transform matrix from registering diagnostic CT to the pCT. We proposed a 3D-UNet-based segmentation method to segment NSCLC tumor targets on dual-modality PET-pCT images. This network contained squeeze-and-excitation and Residual blocks in each convolutional block to perform dynamic channel-wise feature recalibration. Furthermore, up-sampling paths were added to supplement low-resolution features to the model and also to compute the overall loss function. The dice similarity coefficient (DSC), precision, recall, and the average symmetric surface distances were used to assess the performance of the proposed approach on 86 pairs of diagnostic PET and pCT images. The proposed model using dual-modality images was compared with both conventional 3D-UNet architecture and single-modality image input. Results: The average DSC of the proposed model with both PET and pCT images was 0.844, compared to 0.795 and 0.827, when using 3D-UNet and nnUnet. It also outperformed using either pCT or PET alone with the same network, which had DSC of 0.823 and 0.732, respectively. Discussion: Therefore, our proposed segmentation approach is able to outperform the current 3D-UNet network with diagnostic PET and pCT images. The integration of two image modalities helps improve segmentation accuracy.

Segmentation of the cervical lesion region in colposcopic images based on deep learning.

Background: Colposcopy is an important method in the diagnosis of cervical lesions. However, experienced colposcopists are lacking at present, and the training cycle is long. Therefore, the artificial intelligence-based colposcopy-assisted examination has great prospects. In this paper, a cervical lesion segmentation model (CLS-Model) was proposed for cervical lesion region segmentation from colposcopic post-acetic-acid images and accurate segmentation results could provide a good foundation for further research on the classification of the lesion and the selection of biopsy site. Methods: First, the improved Faster Region-convolutional neural network (R-CNN) was used to obtain the cervical region without interference from other tissues or instruments. Afterward, a deep convolutional neural network (CLS-Net) was proposed, which used EfficientNet-B3 to extract the features of the cervical region and used the redesigned atrous spatial pyramid pooling (ASPP) module according to the size of the lesion region and the feature map after subsampling to capture multiscale features. We also used cross-layer feature fusion to achieve fine segmentation of the lesion region. Finally, the segmentation result was mapped to the original image. Results: Experiments showed that on 5455 LSIL+ (including cervical intraepithelial neoplasia and cervical cancer) colposcopic post-acetic-acid images, the accuracy, specificity, sensitivity, and dice coefficient of the proposed model were 93.04%, 96.00%, 74.78%, and 73.71%, respectively, which were all higher than those of the mainstream segmentation model. Conclusion: The CLS-Model proposed in this paper has good performance in the segmentation of cervical lesions in colposcopic post-acetic-acid images and can better assist colposcopists in improving the diagnostic level.

Safety and Efficacy of Stereotactic Ablative Radiotherapy for Ultra-Central Lung Cancer.

Background: Ultra-central lung cancer (UCLC) is difficult to achieve surgical treatment. Over the past few years, stereotactic ablative radiotherapy (SABR) or stereotactic body radiotherapy (SBRT) obviously improved the clinical efficacy and survival of UCLC patients. However, the adapted scheme of radiation therapy is still controversial. For this, a single arm retrospective analysis was performed on UCLC patients treated with SBRT. Material and Methods: We retrospectively studied primary UCLC patients who were treated with SBRT of 56 Gy/6-8f between 2010 and 2018. UCLC was defined as planning target volume (PTV) touching or overlapping the proximal bronchial tree, trachea, esophagus, heart, pulmonary vein, or pulmonary artery within 2 cm around the bronchial tree in all directions. Results: A total of 58 patients whose median age was 68 years (range, 46-85) were included in our study, 79.3% of whom did not undergo any previous therapy. The median dose of the PTV was 77.8 Gy (range, 43.3-91.8), and the median PTV of tumors was 6.2 cm3 (range, 12.9-265.0). With a median follow-up of 57 months (range, 6-90 months), the median cumulative overall survival (OS) rate was 58 months (range, 2-105). In addition, the 1-year, 2-year and 5-year OS rates were 94.7%, 75.0% and 45.0%, respectively. In our univariable analysis (p=0.020) and multivariate analysis (p=0.004), the OS rate was associated with the PTV. The 5-year OS rates for PTV <53.0 cm3 and PTV ≥53.0 cm3 were 61.6% and 37.4%, respectively. Regarding toxicity after SBRT, there were two cases (3.5%) with grade ≥3 adverse events, of which 1 case died of sudden severe unexplained hemoptysis. Conclusions: Patients with UCLC can benefit from SBRT at a dose of 56 Gy/6-8f. On the other hand, smaller PTV was associated with superior outcomes, and the cure difference needs to be validated by prospective comparative trials.

Design of lung nodules segmentation and recognition algorithm based on deep learning.

BACKGROUND: Accurate segmentation and recognition algorithm of lung nodules has great important value of reference for early diagnosis of lung cancer. An algorithm is proposed for 3D CT sequence images in this paper based on 3D Res U-Net segmentation network and 3D ResNet50 classification network. The common convolutional layers in encoding and decoding paths of U-Net are replaced by residual units while the loss function is changed to Dice loss after using cross entropy loss to accelerate network convergence. Since the lung nodules are small and rich in 3D information, the ResNet50 is improved by replacing the 2D convolutional layers with 3D convolutional layers and reducing the sizes of some convolution kernels, 3D ResNet50 network is obtained for the diagnosis of benign and malignant lung nodules. RESULTS: 3D Res U-Net was trained and tested on 1044 CT subcases in the LIDC-IDRI database. The segmentation result shows that the Dice coefficient of 3D Res U-Net is above 0.8 for the segmentation of lung nodules larger than 10 mm in diameter. 3D ResNet50 was trained and tested on 2960 lung nodules in the LIDC-IDRI database. The classification result shows that the diagnostic accuracy of 3D ResNet50 is 87.3% and AUC is 0.907. CONCLUSION: The 3D Res U-Net module improves segmentation performance significantly with the comparison of 3D U-Net model based on residual learning mechanism. 3D Res U-Net can identify small nodules more effectively and improve its segmentation accuracy for large nodules. Compared with the original network, the classification performance of 3D ResNet50 is significantly improved, especially for small benign nodules.

Benefit of dosimetry distribution for patients with multiple brain metastases from non-small cell lung cancer by a Cyberknife stereotactic radiosurgery (SRS) system.

BACKGROUND: In order to obtain a high dose conformal index of tumor and steep dose fall-off in healthy tissues for brain metastasis stereotactic radiosurgery (SRS), the aim of this study was to investigate SRS planning optimization by comparing one multiple-lesions plan (MLP) with multiple single-lesion plans (SLPs) for patients with multiple brain metastases using the Cyberknife (CK) system. METHODS: Fifty non-small cell lung cancer (NSCLC) patients (28 males and 22 females) with 2-4 brain metastases, inter-tumour distances less than 3 cm, were retrospectively replanned with the original prescription dose (12-32 Gy) in the original fractions (1-3). Two different clinical CK SRS plans (SLPs and MLP) were generated for the same patients with the same collimator and prescription isodose line (62-68%) by the CK Multiplan System. Both SLPs and MLP were able to achieve > 95% PTV volume covered prescription dose and met the Timmerman 2011 organs at risk (brainstem, optic nerve and pituitary) constraints. RESULTS: Compared with those in the SLPs, the maximum dose (Dmax) and mean dose (Dmean) of brainstem in the MLP were reduced 0.22-3.13% (2.62%) and 2.71-12.56% (5.57%), respectively, all P < 0.05. Meanwhile, the volumes of the whole brain minus the tumors that received a single dose equivalent of 8-16 Gy (V8Gy-V16Gy) were effectively reduced in the MLP. The treatment time parameters, the total number of beams and monitor units, of the MLP were reduced by 3.31 and 1.47% (P < 0.05), respectively. Although there were a few differences in the conformity index (CI) and homogeneity index (HI) between the two treatment plans, the differences were not statistically significant (P = 2.94 and 1.08 > 0.05). CONCLUSION: One multiple-lesions plan for brain metastases could achieve higher precision in the target and lower doses in healthy tissue while shortening the treatment time and improving the treatment efficiency over multiple single-lesion plans.

Risk-adapted stereotactic body radiation therapy for central and ultra-central early-stage inoperable non-small cell lung cancer.

To determine the therapeutic efficacy and safety of risk-adapted stereotactic body radiation therapy (SBRT) schedules for patients with early-stage central and ultra-central inoperable non-small cell lung cancer. From 2006 to 2015, 80 inoperable T1-2N0M0 NSCLC patients were treated with two median dose levels: 60 Gy in six fractions (range, 48-60 Gy in 4-8 fractions) prescribed to the 74% isodose line (range, 58%-79%) for central lesions (ie within 2 cm of, but not abutting, the proximal bronchial tree; n = 43), and 56 Gy in seven fractions (range, 48-60 Gy in 5-10 fractions) prescribed to the 74% isodose line (range, 60%-80%) for ultra-central lesions (ie abutting the proximal bronchial tree; n = 37) on consecutive days. Primary endpoint was overall survival (OS); secondary endpoints included progression-free survival (PFS), tumor local control rate (LC), and toxicity. Median OS and PFS were 64.47 and 32.10 months (respectively) for ultra-central patients, and not reached for central patients. Median time to local failure, regional failure, and any distant failures for central versus ultra-central lesions were: 27.37 versus 26.07 months, 20.90 versus 12.53 months, and 20.85 versus 15.53 months, respectively, all P < .05. Multivariate analyses showed that tumor categorization (ultra-central) and planning target volume ≥52.76 mL were poor prognostic factors of OS, PFS, and LC, respectively (all P < .05). There was one grade 5 toxicity; all other toxicities were grade 1-2. Our results showed that ultra-central tumors have a poor OS, PFS, and LC compared with central patients because of the use of risk-adapted SBRT schedules that allow for equal and favorable toxicity profiles.

The segmentation of bones in pelvic CT images based on extraction of key frames.

BACKGROUND: Bone segmentation is important in computed tomography (CT) imaging of the pelvis, which assists physicians in the early diagnosis of pelvic injury, in planning operations, and in evaluating the effects of surgical treatment. This study developed a new algorithm for the accurate, fast, and efficient segmentation of the pelvis. METHODS: The proposed method consists of two main parts: the extraction of key frames and the segmentation of pelvic CT images. Key frames were extracted based on pixel difference, mutual information and normalized correlation coefficient. In the pelvis segmentation phase, skeleton extraction from CT images and a marker-based watershed algorithm were combined to segment the pelvis. To meet the requirements of clinical application, physician's judgment is needed. Therefore the proposed methodology is semi-automated. RESULTS: In this paper, 5 sets of CT data were used to test the overlapping area, and 15 CT images were used to determine the average deviation distance. The average overlapping area of the 5 sets was greater than 94%, and the minimum average deviation distance was approximately 0.58 pixels. In addition, the key frame extraction efficiency and the running time of the proposed method were evaluated on 20 sets of CT data. For each set, approximately 13% of the images were selected as key frames, and the average processing time was approximately 2 min (the time for manual marking was not included). CONCLUSIONS: The proposed method is able to achieve accurate, fast, and efficient segmentation of pelvic CT image sequences. Segmentation results not only provide an important reference for early diagnosis and decisions regarding surgical procedures, they also offer more accurate data for medical image registration, recognition and 3D reconstruction.

Preliminary Clinical Validation of a Differential Correction Method for Improving Measurement Accuracy in Noninvasive Measurement of Blood Glucose Using Near-Infrared Spectroscopy.

One of the main challenges in the noninvasive sensing of blood glucose by near-infrared (NIR) spectroscopy is the background variations from light source drift, sweating, and temperature change at the human-machine interface. In this paper, a differential correction method based on the spectra from the floating-reference position and measuring position is proposed to eliminate these spectral variations from background interferences. Its effectiveness was validated by in vitro and in vivo experiments in which the diffuse reflectance of intralipid solutions and human skin was collected at the source distances of 0.6 mm and 2 mm by the custom-built system with six super-luminescent emitting diodes (SLEDs) light source. The results showed that, for the in vitro experiments of intralipid solutions, the coefficients of variations of diffuse reflectance decreased by 20.5% under all the six wavelengths after differential correction. For the in vivo experiments of oral glucose tolerance tests (OGTTs), partial least squares (PLS) regression models between glucose concentrations and the diffuse reflectance from palm skin were built, and the root mean square error of cross validation (RMSECV) decreased by 38.0% on average after the differential correction. Further, the spectra of the oral water tolerance tests (OWTTs) were collected for correlation with glucose concentration in OGTTs, and their correlation coefficients (R) decreased by 35.0% on average after the differential correction. Therefore, this differential correction method based on the spectra from the floating-reference position and measuring position can weaken the influence of background variations on the NIR spectroscopy and has promising potential in in vivo detection, especially for noninvasive blood glucose measurement.

Research on the best measurement situation between optical probe and tissue surfaces in non-invasive detection.

Near-infrared spectroscopy is often used for the non-invasive detection of composition in the human body, such as that of blood glucose and haemoglobin, due to its high penetration depth into tissues. Although it is feasible to position the optical probe precisely, contact situation between probe and human tissues is a difficult problem to determine because of physiological tremor and mechanical performance of bio-soft tissue. Here, we proposed a novel estimation method for the situation between the optical probe and tissue surfaces based on the dynamic auto-correlation matrix of two-dimensional correlation spectroscopy (2DCOS) and radar chart. The diffuse reflectance spectra from the left palm of 4 healthy volunteers were collected while the optical probe gradually approached and pressed bio-tissues with a custom-design controlling device. 2DCOS in the wavelength with lower absorption (1000-1400 nm) was calculated under the perturbation of relative-distance and contact pressure between the optical probe and tissue surface. The synchronous 2DCOS showed that the surface reflection and diffuse reflectance were greatly affected by the contact conditions in 1100 nm, 1220 nm, and 1300 nm. Then the dynamic auto-correlation matrix of 2DCOS was established for the adjacent spectra, and the significant difference wavelengths were used to build radar charts to determine the critical contact situation visually. Results showed that the maximum variations of dynamic auto-correlation matrix appeared at near 1300 nm, and the relative distance between the probe and tissue corresponding to the critical contact state can be easily observed with radar charts with 0.25 mm uncertainty, which was consistent with the self-feeling of each volunteer. So this method can be applied to exactly determine the optimal measurement status for the non-invasive body composition detection in vivo. It is important for the design of human-machine interface and the accuracy improvement of body composition measurements.

[Study on the Effect of Sodium Chloride Salt on Near-Infrared Spectroscopy of Glucose Aqueous Solution].

The sodium chloride (NaCl) salt has been reported to be associated with glucose metabolism. However, the effect of it on non-invasive detection of blood glucose using near-infrared spectroscopy is still an open question. The aim of this study was to investigate this affection through transform background correction analysis two-dimensional (2D) correlation synchronous spectrum and the partial least-squares (PLS) regression. First, the transmittances of glucose aqueous solutions with different NaCl content are collected and the pure water and NaCl aqueous solution are measured as the background. Results show that, the dissolving of NaCl in water changes the amplitude and position of the absorption peak of water. There are two negative peaks in 1 400 and 1 500~1 700 nm corrected spectra of NaCl aqueous obviously and the amplitude of peaks associated with NaCl concentration. That's because NaCl affect the molecular binding and vibration of water. Then the glucose aqueous solutions without NaCl and with NaCl are corrected by the spectra of pure water and NaCl aqueous solution, respectively. So we get the conclusion that NaCl also affect the combination of glucose and water molecules. And the two-dimensional correlation spectroscopy analysis is performed under the perturbation of glucose concentration. The slice spectra of synchronous correlation spectra show that, the adding of NaCl weakens the spectral variation due to glucose concentration change in the wavelength of 1 400 and 1 520~1 700 nm. Finally, the partial least square (PLS) regression models were built to quantitatively conduct the influence of NaCl on glucose prediction accuracy. Comparison results showed that, NaCl molecule in aqueous solution will deteriorate the model accuracy, where root mean square error of prediction increases with the NaCl content; the mean difference of predicted glucose concentration between models based on glucose aqueous solutions with NaCl and without NaCl, is linear with NaCl concentration in samples.

An improved algorithm of temperature compensation for a near infrared multiple-acquisition system based on two-dimensional regression analysis.

The near infrared (NIR) spectroscopy analytical technique is one of the most advanced and promising tools in many domains. NIR acquisition is easily influenced by temperature, thereby affecting qualitative and quantitative analyses. In this paper, a temperature compensation model was established between NIR signals and output voltage values based on two-dimensional regression analysis. The effectiveness of the proposed compensation scheme was experimentally demonstrated by the measurement of six super luminescent diode sources at 293-313 K. The coefficient of variation was decreased 2-fold with this compensation algorithm. The results indicated that it was suitable for various NIR spectral acquisition systems with lower complexity and a higher signal-noise-ratio after being applied to an acousto-optic-tunable-filter system.

Digital human modeling and its applications: Review and future prospects.

OBJECTIVE: The digital human modeling combines information technology with bioscience, applied to analog study from DNA molecule and protein to cell and tissue, as well as organ. It has been widely implemented in various fields such as aviation, national defense, film and television, sports and medical treatment. This paper sorts relative research achievements and progress in recent years and summarizes issues and challenges of research work. The purpose of this paper is to provide references for the development of Digital Human Modeling. METHODS: Basing on the development plans proposed by Federation of American Scientists (FAS) and current theories, we sorted academic achievements and relative references since the beginning of the digital human concept until 2014 into four aspects, which were named as Visible Human, Virtual Physical Human, Virtual Physiological Human, and Intelligent Virtual Human. We collected information of data acquisition, data analysis and dataset establishment for Visible Human. We outlined the Virtual Physical Human references in four physical categories, including radiation, ultrasonic, electrics, and mechanics. For Virtual Physiological Human, relative researches related to the physiological and biochemical changes of human body were categorized in four aspects as gene molecules, cells, and organs. Relative researches of Virtual Brain Human were mainly concentrated on virtual human brain and virtual human control. RESULTS: According to the differences in investigations and applications requirements of Digital Human in all aspects, we classified, analyzed and concluded extensive preference data. The overall plan of Digital Human, as well as technical routes, work plans, present states, goals, results and application value in each stage of research was summarized. We discussed the technical issues existing in each aspect of Digital Human, along with the urgent key technologies including data collection, data processing, and modeling. CONCLUSION: Analyzing the core technologies and the general technical schemes in the field of Digital Human, this paper summarizes the technical solutions, research results and technical problems of major researchers. The future prospects of Digital Human are reviewed.