Deep learning-based denoising algorithm in comparison to iterative reconstruction and filtered back projection: a 12-reader phantom study.

OBJECTIVES: (1) To compare low-contrast detectability of a deep learning-based denoising algorithm (DLA) with ADMIRE and FBP, and (2) to compare image quality parameters of DLA with those of reconstruction methods from two different CT vendors (ADMIRE, IMR, and FBP). MATERIALS AND METHODS: Using abdominal CT images of 100 patients reconstructed via ADMIRE and FBP, we trained DLA by feeding FBP images as input and ADMIRE images as the ground truth. To measure the low-contrast detectability, the randomized repeat scans of Catphan® phantom were performed under various conditions of radiation exposures. Twelve radiologists evaluated the presence/absence of a target on a five-point confidence scale. The multi-reader multi-case area under the receiver operating characteristic curve (AUC) was calculated, and non-inferiority tests were performed. Using American College of Radiology CT accreditation phantom, contrast-to-noise ratio, target transfer function, noise magnitude, and detectability index (d') of DLA, ADMIRE, IMR, and FBPs were computed. RESULTS: The AUC of DLA in low-contrast detectability was non-inferior to that of ADMIRE (p < .001) and superior to that of FBP (p < .001). DLA improved the image quality in terms of all physical measurements compared to FBPs from both CT vendors and showed profiles of physical measurements similar to those of ADMIRE. CONCLUSIONS: The low-contrast detectability of the proposed deep learning-based denoising algorithm was non-inferior to that of ADMIRE and superior to that of FBP. The DLA could successfully improve image quality compared with FBP while showing the similar physical profiles of ADMIRE. KEY POINTS: • Low-contrast detectability in the images denoised using the deep learning algorithm was non-inferior to that in the images reconstructed using standard algorithms. • The proposed deep learning algorithm showed similar profiles of physical measurements to advanced iterative reconstruction algorithm (ADMIRE).

Radiation Dose Reduction in Digital Mammography by Deep-Learning Algorithm Image Reconstruction: A Preliminary Study.

Purpose: To develop a denoising convolutional neural network-based image processing technique and investigate its efficacy in diagnosing breast cancer using low-dose mammography imaging. Materials and Methods: A total of 6 breast radiologists were included in this prospective study. All radiologists independently evaluated low-dose images for lesion detection and rated them for diagnostic quality using a qualitative scale. After application of the denoising network, the same radiologists evaluated lesion detectability and image quality. For clinical application, a consensus on lesion type and localization on preoperative mammographic examinations of breast cancer patients was reached after discussion. Thereafter, coded low-dose, reconstructed full-dose, and full-dose images were presented and assessed in a random order. Results: Lesions on 40% reconstructed full-dose images were better perceived when compared with low-dose images of mastectomy specimens as a reference. In clinical application, as compared to 40% reconstructed images, higher values were given on full-dose images for resolution (p < 0.001); diagnostic quality for calcifications (p < 0.001); and for masses, asymmetry, or architectural distortion (p = 0.037). The 40% reconstructed images showed comparable values to 100% full-dose images for overall quality (p = 0.547), lesion visibility (p = 0.120), and contrast (p = 0.083), without significant differences. Conclusion: Effective denoising and image reconstruction processing techniques can enable breast cancer diagnosis with substantial radiation dose reduction.

Determination of pharmacokinetics and tissue distribution of a novel lutetium-labeled PSMA-targeted ligand, 177Lu-DOTA-PSMA-GUL, in rats by using LC-MS/MS.

Prostate specific membrane antigen (PSMA) is known to be overexpressed in prostate cancer cells, providing as a diagnostic and therapeutic target for prostate cancer. A lutetium-labeled PSMA targeted ligand, 177Lu-DOTA-PSMA-GUL is a novel radiopharmaceutical, which has been developed for the treatment of prostate cancer. While the GUL domain of 177Lu-DOTA-PSMA-GUL binds to the antigen, the beta-emitting radioisotope, 177Lu-labeled DOTA, interacts with prostate cancer cells. However, the in vivo pharmacokinetics of intact 177Lu-DOTA-PSMA-GUL has never been characterized. This study aimed to evaluate the pharmacokinetics and tissue distribution of the radiopharmaceutical in rats by using its stable isotope-labeled analog, 175Lu-DOTA-PSMA-GUL. A sensitive liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of 175Lu-DOTA-PSMA-GUL was developed and validated. Following intravenous injection, the plasma concentration-time profiles of 175Lu-DOTA-PSMA-GUL showed a multi-exponential decline with the average elimination half-life of 0.30 to 0.33 h. Systemic exposure increased with the dose and renal excretion is the major elimination route. Tissue distribution of 175Lu-DOTA-PSMA-GUL was most substantial in kidneys, followed by the prostate. The developed LC-MS/MS assay and the in vivo pharmacokinetic data of 175Lu-DOTA-PSMA-GUL would provide helpful information for further clinical studies to be developed as a novel therapeutic agent for prostate cancer.

The role of Janus kinase 2 (JAK2) activation in pneumococcal EstA protein-induced inflammatory response in RAW 264.7 macrophages.

In a previous study, we demonstrated pneumococcal EstA-induced inflammatory response through NF-κB and MAPK-dependent pathways. Herein, we tested the hypothesis that the Janus kinase 2 (JAK2) activation and associated signaling cascades may also be involved in EstA-induced inflammatory process in RAW 264.7 macrophages. Our immunoblot analysis indicated EstA-induced activation of JAK2, with the phosphorylated protein detected from 1 to 24 h post-stimulation. As type I interferon (IFN) signaling requires the JAK/STAT pathway, we investigated EstA-induced expression of INF-α4 and INF-ß by semi-quantitative and quantitative RT PCR. Our results indicated both concentration- and time-dependent increases in both IFN-α4 and IFN-ß mRNA expression after EstA challenge, with the highest fold-increases observed at 4 h and 6 h post-stimulation for IFN-α4 and IFN-ß mRNA, respectively. Furthermore, we applied a pharmacological approach to demonstrate the effect of JAK2 inhibition on EstA-induced nitric oxide (NO) and pro-inflammatory cytokine production. The JAK2 inhibitor AG-490 reduced significantly (P < 0.05) EstA-induced NO production and the expression of iNOS mRNA in a concentration-dependent manner. Similarly, EstA-induced IL-1ß and IL-6 production and their respective mRNA expression were markedly suppressed by AG-490. However, AG-490 had no inhibitory effect on both mRNA and protein levels of TNF-α. Taken together, we demonstrate that JAK2 activation and IFN I signaling are integral parts of EstA-induced inflammatory process. Further studies will elucidate the interaction of the different signaling pathways, the specific downstream targets of JAK2, the kinetics of cytokine release, and if EstA could induce the pro-inflammatory mediators to the same extent in alveolar macrophages.

Effect of DPSS laser on the shear bond strength of orthodontic brackets.

PURPOSE: To test the bonding of orthodontic brackets to teeth using a diode-pumped solid state (DPSS) laser. METHODS: A total of 60 extracted human teeth were divided randomly into four groups: Group 1 (control)--he brackets were bonded to teeth using the quartz-tungsten-halogen (QTH) light (800 mW/cm2) for 40 seconds; Groups 2-4--the brackets were bonded to teeth using the DPSS laser (500 mW/cm2) for 40 seconds, 20 seconds, and 10 seconds, respectively. The teeth were debonded using shear force in a universal testing machine, and the amount of residual adhesive remaining on each tooth was evaluated. Statistical analysis was carried out for the shear bond strength (SBS) and Adhesive Remnant Index (ARI). RESULTS: The brackets bonded using the DPSS laser for 40 seconds showed the highest mean SBS (13.1 +/- 1.2 MPa) among the groups. Furthermore, the DPSS laser with 10 seconds light-curing could achieve 83% of the mean SBS obtained using the QTH light for 40 seconds. The ARI scores showed no differences among all four groups suggesting a similar failure mode.

Low-Dose Abdominal CT Using a Deep Learning-Based Denoising Algorithm: A Comparison with CT Reconstructed with Filtered Back Projection or Iterative Reconstruction Algorithm.

OBJECTIVE: To compare the image quality of low-dose (LD) computed tomography (CT) obtained using a deep learning-based denoising algorithm (DLA) with LD CT images reconstructed with a filtered back projection (FBP) and advanced modeled iterative reconstruction (ADMIRE). MATERIALS AND METHODS: One hundred routine-dose (RD) abdominal CT studies reconstructed using FBP were used to train the DLA. Simulated CT images were made at dose levels of 13%, 25%, and 50% of the RD (DLA-1, -2, and -3) and reconstructed using FBP. We trained DLAs using the simulated CT images as input data and the RD CT images as ground truth. To test the DLA, the American College of Radiology CT phantom was used together with 18 patients who underwent abdominal LD CT. LD CT images of the phantom and patients were processed using FBP, ADMIRE, and DLAs (LD-FBP, LD-ADMIRE, and LD-DLA images, respectively). To compare the image quality, we measured the noise power spectrum and modulation transfer function (MTF) of phantom images. For patient data, we measured the mean image noise and performed qualitative image analysis. We evaluated the presence of additional artifacts in the LD-DLA images. RESULTS: LD-DLAs achieved lower noise levels than LD-FBP and LD-ADMIRE for both phantom and patient data (all p < 0.001). LD-DLAs trained with a lower radiation dose showed less image noise. However, the MTFs of the LD-DLAs were lower than those of LD-ADMIRE and LD-FBP (all p < 0.001) and decreased with decreasing training image dose. In the qualitative image analysis, the overall image quality of LD-DLAs was best for DLA-3 (50% simulated radiation dose) and not significantly different from LD-ADMIRE. There were no additional artifacts in LD-DLA images. CONCLUSION: DLAs achieved less noise than FBP and ADMIRE in LD CT images, but did not maintain spatial resolution. The DLA trained with 50% simulated radiation dose showed the best overall image quality.

Intraarterially delivered human umbilical cord blood-derived mesenchymal stem cells in canine cerebral ischemia.

The present study examined the effects of human umbilical cord blood-derived mesenchymal stem cells (HUCB-derived MSCs) delivered through the basilar artery in a canine thromboembolic brain ischemia model. Cerebral ischemia was induced through occlusion of the middle cerebral artery by injecting thrombus emboli into 10 beagles. In the HUCBC group (n = 5), 1 x 10(6) HUCB-derived MSCs were transplanted through the basilar artery 1 day after ischemic induction using an endovascular interventional approach. In the control group (n = 5), phosphate-buffered saline (PBS) was injected in the same manner in as the HUCBC group. Upon neurobehavioral examination, earlier recovery was observed in the HUCBC group. The HUCBC group showed a decrease in the infarction volume at 1 week after cerebral ischemic induction, whereas the control group showed an increase in the infarction volume at 1 week, by magnetic resonance image analysis. Transplanted cells had differentiated into neurons and astrocytes and were observed in and around endothelial cells that were positive for von Willebrand factor (vWF). HUCB-derived MSCs expressed neuroprotective factors, such as brain-derived neurotrophic factor (BDNF) and vascular endothelial growth factor (VEGF), at 4 weeks after the transplantation. The transplanted cells demonstrated their efficacy by reducing the infarction lesion volume and through earlier recovery from the neurological deficit. These results suggest that intraarterial transplantation of HUCB-derived MSCs could be useful in clinical treatment of cerebral ischemia.

Cycle-consistent adversarial denoising network for multiphase coronary CT angiography.

PURPOSE: In multiphase coronary CT angiography (CTA), a series of CT images are taken at different levels of radiation dose during the examination. Although this reduces the total radiation dose, the image quality during the low-dose phases is significantly degraded. Recently, deep neural network approaches based on supervised learning technique have demonstrated impressive performance improvement over conventional model-based iterative methods for low-dose CT. However, matched low- and routine-dose CT image pairs are difficult to obtain in multiphase CT. To address this problem, we aim at developing a new deep learning framework. METHOD: We propose an unsupervised learning technique that can remove the noise of the CT images in the low-dose phases by learning from the CT images in the routine dose phases. Although a supervised learning approach is not applicable due to the differences in the underlying heart structure in two phases, the images are closely related in two phases, so we propose a cycle-consistent adversarial denoising network to learn the mapping between the low- and high-dose cardiac phases. RESULTS: Experimental results showed that the proposed method effectively reduces the noise in the low-dose CT image while preserving detailed texture and edge information. Moreover, thanks to the cyclic consistency and identity loss, the proposed network does not create any artificial features that are not present in the input images. Visual grading and quality evaluation also confirm that the proposed method provides significant improvement in diagnostic quality. CONCLUSIONS: The proposed network can learn the image distributions from the routine-dose cardiac phases, which is a big advantage over the existing supervised learning networks that need exactly matched low- and routine-dose CT images. Considering the effectiveness and practicability of the proposed method, we believe that the proposed can be applied for many other CT acquisition protocols.

Development of an improved canine model of percutaneous spinal cord compression injury by balloon catheter.

We developed a minimally invasive canine model of spinal cord injury (SCI). A balloon catheter was inserted into the epidural space via the lumbosacral space, and inflated between L2 and L3 for 30 or 60 min under fluoroscopic guidance. Motor function after SCI was assessed using modified Tarlov scale. All seven dogs showed complete paraplegia after the procedure, neurological problems were evident and the modified Tarlov scores remained at zero after the SCI procedure; no improvement in clinical signs was observed. The dogs underwent 3T MR imaging at 3 days and 1 year after SCI. Histopathologic examinations were conducted at 2 weeks, 12 weeks and 1 year after SCI. In the present study, we described an animal model of minimally invasive spinal cord injury using a balloon catheter without laminectomy under fluoroscopic guidance. And, this percutaneous spinal cord compression injury model has many potential applications. The described percutaneous spinal cord compression injury model offers a new means of administering SCI and has many potential applications.

Comparative pharmacokinetics of tylosin or florfenicol after a single intramuscular administration at two different doses of tylosin-florfenicol combination in pigs.

Clinical pharmacokinetic profiles were investigated following intramuscular (i.m.) administration to pigs with a commercial tylosin-florfenicol combination product at a dose of 2.5 mg/kg tylosin and 5 mg/kg florfenicol or 10 mg/kg tylosin and 20 mg/kg florfenicol. The quantitation limit (QL) of florfenicol was 0.1 microg/ml, the inter-day and intra-day precision (CV%) were both beow 10%. The quantitation limit (QL) of tylosin was 0.05 microg/mL. The pharmacokinetic characteristics after i.m. doses were fitted by a one compartment open model. A fourfold decrease in the normal dose of each drug (20 mg/kg to 5 mg/kg for florfenicol, and 10 mg/kg to 2.5 mg/kg for tylosin) resulted in a corresponding two fold decrease in each drug of the maximum plasma concentration (C(max)) and the area under curve (AUC) values.

Corrosion-related changes on Ti-based orthodontic brackets in acetic NaF solutions: surface morphology, microhardness, and element release.

This study sought to investigate the effects of acetic NaF solutions on titanium and Ti alloy brackets. To this end, two different brackets were immersed in various NaF-containing solutions for three days. The Equilibrium Ti (EQ) bracket was composed of Ti only, whereas the Ortho 2 (OR) bracket was composed of Ti (base) and Ti-6A1-4V (wings). Brackets that were immersed in the acetic NaF solution of pH 3.5 yielded no reliable surface microhardness values due to corrosion. In other test solutions, however, there was minimal reduction (at best 3%) in microhardness. Further on microhardness, the values of the OR bracket at the base and wings were different. On the release of elements, it was significant only in the acetic NaF solution of pH 3.5. However, the release of Al (6.11+/-0.93 ppm) and V (1.16+/-0.40 ppm) in this solution was low. In conclusion, an acetic NaF solution of low pH could damage Ti-based orthodontic brackets.

Deep Convolutional Framelet Denosing for Low-Dose CT via Wavelet Residual Network.

Model-based iterative reconstruction algorithms for low-dose X-ray computed tomography (CT) are computationally expensive. To address this problem, we recently proposed a deep convolutional neural network (CNN) for low-dose X-ray CT and won the second place in 2016 AAPM Low-Dose CT Grand Challenge. However, some of the textures were not fully recovered. To address this problem, here we propose a novel framelet-based denoising algorithm using wavelet residual network which synergistically combines the expressive power of deep learning and the performance guarantee from the framelet-based denoising algorithms. The new algorithms were inspired by the recent interpretation of the deep CNN as a cascaded convolution framelet signal representation. Extensive experimental results confirm that the proposed networks have significantly improved performance and preserve the detail texture of the original images.

A deep convolutional neural network using directional wavelets for low-dose X-ray CT reconstruction.

PURPOSE: Due to the potential risk of inducing cancer, radiation exposure by X-ray CT devices should be reduced for routine patient scanning. However, in low-dose X-ray CT, severe artifacts typically occur due to photon starvation, beam hardening, and other causes, all of which decrease the reliability of the diagnosis. Thus, a high-quality reconstruction method from low-dose X-ray CT data has become a major research topic in the CT community. Conventional model-based de-noising approaches are, however, computationally very expensive, and image-domain de-noising approaches cannot readily remove CT-specific noise patterns. To tackle these problems, we want to develop a new low-dose X-ray CT algorithm based on a deep-learning approach. METHOD: We propose an algorithm which uses a deep convolutional neural network (CNN) which is applied to the wavelet transform coefficients of low-dose CT images. More specifically, using a directional wavelet transform to extract the directional component of artifacts and exploit the intra- and inter- band correlations, our deep network can effectively suppress CT-specific noise. In addition, our CNN is designed with a residual learning architecture for faster network training and better performance. RESULTS: Experimental results confirm that the proposed algorithm effectively removes complex noise patterns from CT images derived from a reduced X-ray dose. In addition, we show that the wavelet-domain CNN is efficient when used to remove noise from low-dose CT compared to existing approaches. Our results were rigorously evaluated by several radiologists at the Mayo Clinic and won second place at the 2016 "Low-Dose CT Grand Challenge." CONCLUSIONS: To the best of our knowledge, this work is the first deep-learning architecture for low-dose CT reconstruction which has been rigorously evaluated and proven to be effective. In addition, the proposed algorithm, in contrast to existing model-based iterative reconstruction (MBIR) methods, has considerable potential to benefit from large data sets. Therefore, we believe that the proposed algorithm opens a new direction in the area of low-dose CT research.

Three-dimensional changes in the temporomandibular joint after maxillary protraction in children with skeletal Class III malocclusion.

We evaluated 3-dimensional changes in the temporomandibular joints of children with skeletal Class III malocclusion and maxillary deficiency after facemask therapy for maxillary protraction. Eighteen children with anterior crossbite and a Class III molar relationship underwent facemask therapy for maxillary protraction, after which they exhibited positive overjet and a Class II molar relationship. Three-dimensional cone-beam computed tomography images of the patients were obtained before (T1) and after (T2) facemask protraction, and the 3-dimensional coordinates of the anatomical landmarks in T1 and T2 images were compared. After facemask therapy, the mandibular condyles of the patients were displaced outside, upward, and backward. Additionally, the anterior and posterior walls of the glenoid fossa had negative values for anteroposterior change. Three-dimensional analysis of the temporomandibular joint showed that facemask therapy resulted in bone apposition (to the anterior wall) and bone resorption (of the posterior wall) in the glenoid fossa. This bone remodeling resulted in upward and backward displacement of the condyle.(J Oral Sci 58, 501-508, 2016).

Numerical path integration technique for the calculation of transport properties of proteins.

We present a new technique for the computation of both the translational diffusivity and the intrinsic viscosity of macromolecules, and apply it here to proteins. Traditional techniques employ finite element representations of the surface of the macromolecule, taking the surface to be a union of spheres or of polygons, and have computation times that are O(m(3)) where m is the number of finite elements. The new technique, a numerical path integration method, has computation times that are only O(m). We have applied the technique to approximately 1000 different protein structures. The computed translational diffusivities and intrinsic viscosities are, to lowest order, proportional respectively to N(-1/3)(R) and N(0)(R), where N(R) is the number of amino acid residues in the protein. Our calculations also show some correlation with the shape of the molecule, as represented by the ratio m(2)/m(3), where m(2) and m(3) are, respectively, the middle and the smallest of the three principal moments of inertia. Comparisons with a number of experimental results are also performed, with results generally consistent to within experimental error.

Interspecies comparison of the oral absorption of itraconazole in laboratory animals.

The oral absorption and disposition of itraconazole were studied in rats, rabbits and dogs. Serum levels of itraconazole and its active metabolite, hydroxyitraconazole, were determined by a validated HPLC method. The absorption of itraconazole was relatively rapid in rats and dogs but was slower in rabbits. The terminal elimination half-life (T 1/2lambda(z)), time to the peak concentration (Tmax), dose and weight normalized area under the curve (AUC) and the peak concentration (Cmax) of itraconazole found in the dog were comparable to those reported in humans. As in humans, the metabolite to parent drug AUC ratios in rats and dogs were greater than unity but was less in rabbits. The dog appears to be an appropriate animal model while the rat, not the rabbit, may be used as an alternative animal model in predicting the oral absorption of itraconazole in humans.

Primary mediastinal large B-cell lymphoma: a single-center experience in Korea.

BACKGROUND: Primary mediastinal large B-cell lymphoma (PMBL) is a distinct subtype of non-Hodgkin lymphoma, which has no consensus for its ideal treatment or prognosis. METHODS: We reviewed the clinicopathologic features and clinical outcomes of 25 PMBL cases diagnosed at a single institution between 1993 and 2009 and compared them with 588 cases of non-mediastinal, diffuse large B-cell lymphoma (DLBCL, control group) diagnosed during the same period. RESULTS: Thirteen (52.0%) PMBL patients had Ann Arbor stage III or IV disease, and 10 (40.0%) had B symptoms. Thirteen (52%) PMBL patients were classified as high-intermediate/high-risk according to the International Prognostic Index. There was a significant prevalence of young (median: 31 years; range, 15-78 years; P<0.001), female (68%; P=0.014) patients in the PMBL group compared to the control group (median: 56 years; range, 15-85 years; 43.2% female). Bulky disease and elevated levels of lactate dehydrogenase (LDH) were more frequent in the PMBL group (P<0.001 and P=0.003, respectively). Nineteen (76%) PBML patients achieved complete remission, and 18 were alive at the last follow-up (median: 43 months; range, 1-92 months). There was no difference in the 3-year, overall survival rate (72%, 95% confidence interval [CI]: 54.0-83.0 versus 70.1%, 95% CI, 109.0-126.0; P=0.686) between PMBL and control patients, respectively. CONCLUSION: Compared to patients with non-mediastinal DLBCL, Korean patients with PMBL are predominantly young women with bulky disease and high LDH levels but with no significant difference in survival.

Percutaneous transplantation of human umbilical cord-derived mesenchymal stem cells in a dog suspected to have fibrocartilaginous embolic myelopathy.

The use of human umbilical cord blood-derived mesenchymal stem cells for cell transplantation therapy holds great promise for repairing spinal cord injury. Here we report the first clinical trial transplantation of human umbilical cord (hUCB)-derived mesenchymal stem cells (MSCs) into the spinal cord of a dog suspected to have fibrocartilaginous embolic myelopathy (FCEM) and that experienced a loss of deep pain sensation. Locomotor functions improved following transplantation in a dog. Based on our findings, we suggest that transplantation of hUCB-derived MSCs will have beneficial therapeutic effects on FCEM patients lacking deep pain sensation.

Schwann cell-like remyelination following transplantation of human umbilical cord blood (hUCB)-derived mesenchymal stem cells in dogs with acute spinal cord injury.

Human umbilical cord blood derived mesenchymal stem cells (hUCB-MSCs) have significant therapeutic potential in cell-based therapies following spinal cord injury (SCI). To evaluate this potential, we conducted our preliminary investigations on the remyelination of injured spinal cords with hUCB-MSC transplantations and we observed its long term effects on dogs with SCI. Of the ten injured dogs, seven were transplanted with hUCB-MSCs 1 week after SCI, whereas the remaining three dogs were not transplanted. Two transplanted dogs died over the first month after transplantation because of urinary tract infection, bedsores and sepsis. The SCI dogs showed no improvement in motor and sensory functions and their urinary dysfunction persisted until they were euthanized (from 3 months to 1 year) while hind-limb recovery in 4 dogs among the five transplanted dogs was significantly improved. In the recovered dogs, functional recovery was sustained for three years following transplantation. Histological results from five transplanted dogs showed that many axons were remyelinated by P0-positive myelin sheaths after transplantation. Our results suggest that transplantation of hUCB-derived MSCs may have beneficial therapeutic effects. Furthermore, histological results provided the first in vivo evidence that hUCB-MSCs are able to enhance the remyelination of peripheral-type myelin sheaths following SCI.

Design, synthesis, and X-ray crystallographic analysis of a novel class of HIV-1 protease inhibitors.

In the present paper, design, synthesis, X-ray crystallographic analysis, and HIV-1 protease inhibitory activities of a novel class of compounds are disclosed. Compounds 28-30, 32, 35, and 40 were synthesized and found to be inhibitors of the HIV-1 protease. The crucial step in their synthesis involved an unusual endo radical cyclization process. Absolute stereochemistry of the three asymmetric centers in the above compounds have been established to be (4S,2'R,3'S) for optimal potency. X-ray crystallographic analysis has been used to determine the binding mode of the inhibitors to the HIV-1 protease.